Random Thoughts from a Nonlinear Mind

The All-Star Who (Initially) Did Not Look Like One

Did Travis Bazzana Really Deserve His 2026 Selection?

When Travis Bazzana was named to the 2026 American League All-Star team, my first reaction was surprise. I mean, I was shocked. I didn’t expect him to make the team, did you?

Bazzana had certainly been good. But an All-Star already? He had not even reached the major leagues until April 28. By the time the All-Star roster was selected, he had played only 58 games and accumulated 249 plate appearances.

That made the selection worth examining. I really want to know how and why he made the team.

There was another reason to look closely. Bazzana was not simply added by Major League Baseball to make sure Cleveland had a representative. MLB’s official roster announcement specifically identified him as the American League’s second baseman selected through the Player Ballot, a vote involving players, managers, and coaches. Ernie Clement of Toronto had already been elected the starting second baseman by the fans.

So this was not merely a question of whether Bazzana was good. The more interesting question was whether the players themselves got it right.

Defining “Deserved”

There are several subtly different ways to ask whether someone deserved to be an All-Star. The first is simple: Was Bazzana playing at an All-Star level?

The second is more demanding: Was Bazzana one of the best American League second basemen at the time the roster was selected?

And the third is the hardest: Was he the most deserving reserve once Clement had already been elected the starter?

Those questions do not necessarily produce the same answer. A player can be performing at an All-Star level without being the single best choice available. I’ll take a look to see if he was the best choice.

For this analysis, I used the FanGraphs data available through July 4, immediately before the roster announcement. Clearly, using August statistics to judge a July decision would allow information that voters could not have possessed at the time to contaminate the analysis.

I also imposed a 150-plate-appearance minimum. That eliminated tiny samples while comfortably retaining Bazzana’s 249 plate appearances.

One player required special treatment. Ezequiel Duran did not appear in the FanGraphs second-base-filter export, but he was certainly an official AL second-base candidate on MLB’s ballot and appeared in the broader qualified American League FanGraphs exports. MLB’s June ballot updates even showed him among the leaders at the position. I therefore restored him to the comparison pool.

The First Surprise: Bazzana Was Good

The raw FanGraphs numbers immediately undermine the argument that Bazzana was a reputation-driven selection.

He was slashing:

.250/.341/.412

with 7 home runs, 12 stolen bases, an 11.6 percent walk rate, and a 20.9 percent strikeout rate.

More importantly, his 114 wRC+ indicates that his offensive production was roughly 14 percent better than league average after the adjustments incorporated into the metric. He had accumulated +6.15 offensive runs, +2.07 baserunning runs, and 1.43 WAR in only 249 plate appearances. Those are legitimate numbers.

They do not immediately prove that Bazzana should have been the reserve second baseman. But they establish something important at the outset. The selection was not absurd.

Table 1. Leading AL Second-Base Candidates Through July 4, 2026

Player	PA	wRC+	BsR	Def	WAR
Ezequiel Duran	297	107	+0.88	+8.00	2.22
Jazz Chisholm Jr.	334	98	+4.67	+4.83	2.10
Chase Meidroth	364	104	-0.29	+4.36	1.90
Kody Clemens	301	123	+0.68	-2.22	1.76
Travis Bazzana	249	114	+2.07	-0.91	1.43
Ernie Clement	342	108	-0.60	-1.08	1.38
Cole Young	357	107	-0.81	-1.37	1.37
Gleyber Torres	190	128	-2.91	-0.18	1.01

The FanGraphs exports place Bazzana behind Duran, Jazz Chisholm Jr., Chase Meidroth, and Kody Clemens in total WAR, but slightly ahead of Clement and Cole Young. Duran’s broader AL export shows 2.22 WAR, driven in considerable part by a very strong defensive rating.

Figure 1 changes the tone of the discussion. Figure 1 shows the FanGraphs WAR totals for the leading American League second-base candidates through July 4. The important point is not simply that Bazzana ranked fifth. He remained within the main cluster of serious candidates despite having considerably fewer plate appearances than most of the players ahead of him.

Bazzana was not the WAR leader. He was not even particularly close to Duran. Yet neither was he buried among mediocre players. He occupied the middle of a fairly compact cluster of serious candidates.

That is our first important result.

The Offensive Case for Bazzana

If we look only at offense, Bazzana’s case becomes considerably stronger.

Kody Clemens had the best combination of power and overall offensive production among the leading candidates, carrying a 123 wRC+ and .500 slugging percentage. Gleyber Torres produced an even higher 128 wRC+, although he had only 190 plate appearances and lost substantial value on the bases.

Bazzana sat immediately behind that offensive tier.

His .341 on-base percentage was particularly valuable. He was walking frequently, making enough contact, adding modest power, and creating additional value with his legs.

FanGraphs decomposes offensive value in a useful way:

\mathrm{Offense} = \mathrm{Batting} + \mathrm{Base\ Running}

For Bazzana:

\mathrm{Offense}_{\mathrm{Bazzana}} = 4.076 + 2.074 = 6.150

That is a strong total for someone with only 249 plate appearances. FanGraphs credited him with roughly 4.1 batting runs and another 2.1 runs on the bases.

The baserunning component should not be dismissed as decorative. Twelve stolen bases in 58 games certainly attracted attention, but BsR attempts to measure more than stolen bases alone. Bazzana was producing real value outside the batter’s box.

This is where his All-Star case begins to make intuitive sense. Players facing Cleveland were not necessarily thinking about Bazzana’s WAR ranking. They were seeing a hitter who controlled the strike zone, got on base, ran well, and was already producing above-average offense only weeks into his major-league career.

The Player Ballot may have been capturing something real.

Then Defense Changes Everything

The reason Bazzana’s total WAR was not higher was defense. FanGraphs credited him with -0.91 Def through the cutoff date. That figure combines fielding performance with positional value.

Conceptually:

\mathrm{Defense} = \mathrm{Fielding} + \mathrm{Positional}

For Bazzana, the underlying FanGraphs components were approximately -1.27 fielding runs and +0.36 positional runs, producing the -0.91 total. Compare that with Duran. Duran had only a 107 wRC+, lower than Bazzana’s 114. His total offensive value was +3.31 runs, also well below Bazzana’s +6.15.

But Duran’s FanGraphs defense was an extraordinary +8.00 runs. That defensive advantage propelled him to 2.22 WAR.

Jazz Chisholm Jr. followed a similar path. His 98 wRC+ was actually below league average, yet his baserunning and defense were excellent. FanGraphs gave him +4.67 BsR and +4.83 Def, enough to reach 2.10 WAR.

Kody Clemens was almost the mirror image. His offense was excellent, but FanGraphs rated his defense at -2.22 runs.

That produces an unusually interesting positional race.

Figure 2 plots offensive production against FanGraphs defensive value and shows just how different the candidates were. Bazzana sits on the offense-heavy side of the group, while Duran, Chisholm, and Meidroth received substantially more help from defense.

There was no single prototype for a valuable American League second baseman in early 2026.

Duran and Chisholm were being pushed upward by defense. Kody Clemens was being driven by his bat. Bazzana was getting most of his value from batting and baserunning while surrendering a relatively small amount defensively.

Figure 3 makes that contrast even clearer. The same contrast becomes more evident when the components are separated. Figure 3 compares batting, baserunning, and defensive value for the principal candidates. Bazzana’s profile is distinctive: strong batting value, meaningful baserunning value, and a relatively small defensive penalty.

This is also why defense must be included in the analysis. Ignoring it would artificially elevate Bazzana and Kody Clemens while punishing Duran, Chisholm, and Meidroth.

At the same time, there is a legitimate reason not to treat 80 games of defensive measurement as absolute truth. Defensive statistics stabilize slowly. Positioning, opportunity, scoring systems, and relatively small numbers of fielding plays can substantially affect estimates.

That creates a useful sensitivity test.

What Happens If We Trust Defense Less?

Instead of pretending that defensive measurement is perfectly precise, we can ask how the ranking changes as we gradually reduce its influence.

FanGraphs’ Runs Above Replacement framework can be represented approximately as:

\mathrm{RAR} = \mathrm{Offense} + \mathrm{Defense} + \mathrm{League} + \mathrm{Replacement}

WAR then converts those runs into wins:

\mathrm{WAR} = \frac{ \mathrm{RAR} }{ \mathrm{Runs\ per\ Win} }

For a sensitivity check, I recalculated the candidates while allowing defense to receive 0, 0.5, or its full FanGraphs weight. This is not a replacement version of WAR. It is simply a way to see how dependent the ranking is on defensive evaluation. Because defensive measurements are especially uncertain over partial seasons, it is worth asking whether the conclusion depends heavily on them. Figure 4 performs that sensitivity test by recalculating the candidates’ diagnostic value with defense given zero weight, half weight, and its full FanGraphs weight.

The result is revealing. When defense receives its full FanGraphs weight, Duran leads at 2.22 WAR, followed by Chisholm and Meidroth. Bazzana sits at 1.43.

When defense is given only half weight, Duran falls back toward the pack. Kody Clemens and Chisholm move near the top, while Bazzana remains around 1.48.

When defense is removed entirely, Kody Clemens becomes the leader and Bazzana rises to roughly third among the principal candidates.

Bazzana’s position is unusually stable. He does not need a favorable defensive estimate to make his case. Quite the opposite. Defense is holding his total down.

Duran’s candidacy is much more sensitive. His argument for being clearly superior to Bazzana rests heavily on trusting the defensive estimate.

That does not mean we should throw defense away. It means the gap between Duran and Bazzana is less certain than the raw 2.22 versus 1.43 WAR comparison initially suggests.

The Playing-Time Problem

There is another issue working against Bazzana. He simply had fewer opportunities.

Duran had 297 plate appearances. Chisholm had 334. Meidroth had 364. Clement had 342. Cole Young had 357.

Bazzana had 249.

That was not because he had been ineffective or injured for much of the major-league season. He did not make his debut until April 28.

One simple way to examine the effect of playing time is to normalize WAR to 600 plate appearances:

\mathrm{WAR}_{600} = 600 \left( \frac{ \mathrm{WAR} }{ \mathrm{PA} } \right)

For Bazzana:

\mathrm{WAR}_{600,\mathrm{Bazzana}} = 600 \left( \frac{ 1.432 }{ 249 } \right) \approx 3.45

That is not a projection. It does not mean Bazzana would necessarily have finished a full season with 3.45 WAR per 600 plate appearances.

It simply normalizes the rate at which value had accumulated. By that measure, Bazzana’s performance looks considerably more All-Star-like. His roughly 3.45 WAR-per-600 pace was competitive with the best players in the group.

The tradeoff is philosophical. Should an All-Star ballot reward the player who has accumulated the most value, or the player who has performed at the highest level while actually on the field?

There is no universally correct answer. If total contribution is the standard, Bazzana’s late arrival hurts him. If, on the other hand, quality of performance matters heavily, his case improves.

The Statcast Warning

There is, however, one piece of evidence that prevents me from becoming completely enthusiastic about Bazzana’s first-half numbers. His actual production was noticeably better than his Statcast expected production.

FanGraphs reported:

wOBA: .3319

xwOBA: .3025

SLG: .412

xSLG: .352

AVG: .250

xBA: .228

His barrel rate was only 4.2 percent, and his hard-hit rate was 36.1 percent.

We can express the difference between actual and expected wOBA as:

\Delta \mathrm{wOBA} = \mathrm{wOBA} - \mathrm{xwOBA}

For Bazzana:

\Delta \mathrm{wOBA}_{\mathrm{Bazzana}} = 0.3319 - 0.3025 = 0.0294

That is not trivial. It suggests that the quality of Bazzana’s batted-ball contact did not fully account for the results he produced.

But here again, context matters. Bazzana was not uniquely fortunate. A look at other players shows that Ernie Clement’s actual wOBA was about .324 while his xwOBA was only .272, an even larger gap. Duran also exceeded his xwOBA, although by a smaller amount, posting roughly .321 versus .303.

So Statcast weakens Bazzana’s case somewhat, but it does not destroy it. If anything, it warns us against treating the first-half offensive results of several second basemen as perfectly stable indicators of true talent. Figure 5 compares actual wOBA with Statcast expected wOBA for the second-base candidates. Players above the diagonal performed better than their contact quality would predict, while those below it underperformed their expected results. Bazzana is noticeably above the line.

The Duran Problem

At this point, one player becomes impossible to ignore, Ezequiel Duran.

If the question is simply, Who had accumulated the most FanGraphs value among the legitimate AL second-base candidates?, Duran has the cleanest answer.

He had 2.22 WAR, more plate appearances than Bazzana, and his offense was above average. And FanGraphs credited him with roughly eight defensive runs above average.
He was not an obscure candidate discovered only after the fact. MLB’s own All-Star voting updates showed Duran running near the top of the second-base voting while Bazzana was also among the leading candidates.

If I were selecting the reserve strictly from the FanGraphs value totals available at the time, I would choose Duran over Bazzana. But I would attach an asterisk to that conclusion. The size of Duran’s advantage depends heavily on defense, and half-season defensive values carry more uncertainty than half-season batting statistics. Take away the defensive separation, and Bazzana suddenly looks extremely competitive.

That makes the decision closer than the raw WAR totals imply.

And What About Clement?

There is also an interesting irony in the comparison with Ernie Clement. Clement was not merely elected to start. He became the leading American League vote-getter in Phase 1, automatically locking down the starting assignment at second base.

Statistically, however, Bazzana’s FanGraphs case was at least as strong.

Clement had 1.38 WAR to Bazzana’s 1.43. He had a 108 wRC+ to Bazzana’s 114. His baserunning contribution was negative while Bazzana’s was strongly positive. Both carried slightly negative FanGraphs defensive values in this snapshot.

Clement did have the playing-time advantage, 342 plate appearances compared with Bazzana’s 249. But if someone accepts Clement as a legitimate All-Star starter on performance grounds, it becomes difficult to argue that Bazzana was nowhere near All-Star caliber.

That comparison strengthens Bazzana’s defense considerably.

What Were the Players Seeing?

Statistics cannot tell us why players, managers, and coaches voted for Bazzana, but they allow us to construct a plausible explanation.

He was a rookie who had almost immediately become an above-average major-league hitter. He controlled the strike zone. He reached base. He ran aggressively and effectively. He was producing roughly 14 percent better offense than league average, and he had accumulated 1.4 WAR despite missing the opening month of the major-league season.

Players may also perceive certain skills more directly than statistical models do. They see bat speed, pitch recognition, baserunning pressure, defensive positioning, approach, and the quality of at-bats firsthand.

Reputation may have helped Bazzana. Being the first overall draft pick certainly made him familiar. But reputation alone is not a satisfactory explanation for the vote.

The numbers were already substantial enough to support it.

So, Did Travis Bazzana Deserve to Be an All-Star?

After reviewing the data, I would split the verdict into two parts. Did Travis Bazzana perform well enough to deserve serious All-Star consideration? Yes.

That conclusion is stronger than I expected when I began the analysis. A 114 wRC+, .341 OBP, positive baserunning value, and 1.43 WAR in only 249 plate appearances is an All-Star-caliber performance over the time he actually played.

His selection was not some inexplicable triumph of hype over production.

But the second question produces a different answer. Was Travis Bazzana clearly the most deserving reserve second baseman in the American League? No.

Ezequiel Duran had the strongest FanGraphs total-value case. Jazz Chisholm Jr., Chase Meidroth, and Kody Clemens also had legitimate statistical arguments. Bazzana’s relatively limited playing time and slightly negative defensive value keep him from being the obvious choice.

If I had been required to select one reserve strictly on the performance data available through July 4, I probably would have selected Duran.

Still, “probably Duran” is very different from “Bazzana did not deserve it.”

That distinction is the central finding of this exercise.

A Better Verdict

My original reaction to Bazzana’s selection was essentially: Really? Already?

The data changed that reaction. The best description of the selection is that it was not undeserved. Nor is it obvious but it is defensible.

Bazzana was already one of the strongest offensive second basemen in the American League. He added meaningful value on the bases. His WAR total was competitive despite substantially fewer plate appearances than most of the other leading candidates. Defense weakened his candidacy, and his Statcast-expected numbers suggest that some of his offensive production exceeded the quality of his contact.

Those are real reservations. But they are reservations about whether he was the best choice, not whether he belonged in the conversation.

That may be the most interesting thing about All-Star selections. The statistics often do not identify one incontrovertible winner. They reveal the tradeoffs.

Duran offered accumulated value and defense.

Chisholm offered defense and speed.

Kody Clemens offered offensive power.

Meidroth offered a balanced package with substantial defensive value.

Bazzana offered offense, on-base ability, speed, and an impressive rate of value accumulation in a shortened first half.

The Player Ballot chose Bazzana. The numbers say that choice was arguable yet reasonable.

My guess is that there will be many more All-Star games in Bazzana’s future. He is an impressive young player.

 

How Many Pitching Statistics Do We Really Need?

Correlation, Redundancy, and the Search for Independent Information in Baseball’s Pitching Metrics

Baseball has no shortage of pitching statistics.

ERA is still with us. So is WHIP. But they now share the stage with FIP, xFIP, SIERA, K-BB%, WAR, BABIP, ground-ball percentage, strikeout percentage, walk percentage, home-run rate, and an expanding collection of increasingly specialized measures.

That creates an interesting problem.

Are all these statistics actually telling us different things?

Or have we created many different ways of describing the same underlying pitching abilities?

The question became especially interesting after my study of WHIP. WHIP was very strongly associated with same-season ERA, but it was considerably weaker at predicting ERA one year later. FIP, xFIP, SIERA, and K-BB% all performed better as forward-looking measures.

That suggested a different question.

Perhaps we should stop asking which statistic is “best.”

Instead, we should ask:

How much independent information does each statistic actually contain?

That is the question I investigate here.

The Data

I used the same season-level FanGraphs dataset covering 2002 through 2025.

The data include season, innings pitched, ERA, FIP, xFIP, WAR, BABIP, home-run rate, and related pitching measures. A separate rate-statistics export includes WHIP, K%, BB%, K-BB%, ERA-, FIP-, xFIP-, FIP, xFIP, and SIERA. The batted-ball data add ground-ball percentage and related contact measures.

As in the WHIP study, I required at least 100 innings in a season:

IP_{i,y} \geq 100

That produced 1,699 qualifying pitcher-seasons.

For the predictive analysis, a pitcher had to reach 100 innings in consecutive seasons:

IP_{i,y} \geq 100 \quad \text{and} \quad IP_{i,y+1} \geq 100

That left 927 consecutive-season pairs involving 299 different pitchers.

Because no pitcher reached 100 innings during the shortened 2020 season, that year naturally drops out of the consecutive-season analysis.

The Metrics

The main study included:

ERA, WHIP, FIP, xFIP, SIERA, K%, BB%, K-BB%, HR/9, BABIP, GB%, and WAR.

Some of these variables are obviously related.

One relationship is exact:

\mathrm{K\!-\!BB\%} = K\% - BB\%

This simple equation illustrates the broader issue surprisingly well.

K%, BB%, and K-BB% may occupy three separate columns on a leaderboard, but they do not represent three independent pieces of information.

K-BB% is constructed directly from the other two.

The relationships among FIP, xFIP, SIERA, ERA, WHIP, and the underlying pitching rates are more complicated.

But the same basic problem remains.

Different names do not necessarily mean different information.

First Look: The Correlation Matrix

The natural place to begin is with Pearson correlation.

Rather than displaying the variables in an arbitrary order, I used hierarchical clustering to place statistics with similar correlation structures near one another.

Figure 1. Correlation Structure of Major Pitching Metrics

Several relationships immediately stand out.

Metrics	Correlation
xFIP and SIERA	0.970
K% and K-BB%	0.939
FIP and WAR	-0.885
FIP and xFIP	0.873
SIERA and K-BB%	-0.858
FIP and SIERA	0.854
ERA and WHIP	0.811

The relationship between xFIP and SIERA is extraordinary.

Their correlation is approximately:

r_{\mathrm{xFIP},\mathrm{SIERA}} = 0.970

Squaring that correlation gives:

R^2 = (0.970)^2 \approx 0.941

So roughly 94 percent of their observed variation is shared in a simple linear sense.

That does not make xFIP and SIERA identical. They are calculated differently and emphasize somewhat different aspects of pitching.

But statistically, they move together to an extraordinary degree.

ERA and WHIP tell a similar, though less extreme, story. Their same-season correlation is approximately 0.81.

WHIP and ERA look like different statistics.

In practice, they often move together.

Correlation Is Only the Beginning

Pairwise correlation cannot tell us everything.

A statistic may have only moderate correlations with several individual variables while still being highly predictable from all of them collectively.

This matters enormously in regression analysis.

Suppose we try to predict one variable using all the others. If that variable can already be reconstructed very accurately, then adding it to a large regression may provide very little truly new information.

A standard way to investigate this is the Variance Inflation Factor:

\mathrm{VIF}_j = \frac{ 1 }{ 1-R_j^2 }

Here, R_j^2 measures how well predictor (j) can itself be predicted by the other predictors.

A VIF near 1 suggests relatively little redundancy.

As VIF increases, multicollinearity becomes more serious.

I excluded K-BB% from this particular calculation because K%, BB%, and K-BB% have an exact mathematical dependency.

The remaining results were striking.

Figure 2. Multicollinearity Among the Pitching Metrics

Metric	VIF
FIP	109.0
WHIP	93.1
K%	47.5
HR/9	46.9
xFIP	44.1
BB%	32.1
BABIP	29.2
SIERA	28.2
ERA	5.8
GB%	3.9

These values are enormous.

But they should not be interpreted as evidence that FIP, WHIP, or SIERA are bad statistics.

That is not what VIF measures.

The result, instead, shows that combining all these variables into a single regression equation creates extreme redundancy.

Several predictors are trying to explain the same underlying variation.

That will matter shortly.

What Should We Predict?

A statistic can look extremely impressive when it is asked to explain something happening in the same season.

Prediction is harder.

So, as in the WHIP study, I used next-season ERA as the principal target.

The general idea is:

\widehat{\mathrm{ERA}}_{i,y+1} = \beta_0 + \sum_{j=1}^{p} \beta_j z_{i,y,j}

where the predictors come from year (y), while the target is ERA in year (y+1).

The predictors were standardized:

z_{i,y,j} = \frac{ x_{i,y,j} - \overline{x}_{j} }{ s_j }

Standardization allows coefficients from variables measured on very different numerical scales to be compared more sensibly.

More importantly, I did not simply fit the models and report their in-sample (R²).

I used leave-one-season-out cross-validation.

One season was withheld.

The model was trained using all the other seasons.

Then it had to predict the observations belonging to the season it had not seen.

The procedure was repeated across the available seasons.

That creates a much more demanding test.

Which Single Metric Predicts Best?

Before constructing complicated regression models, it makes sense to give each statistic a chance by itself.

Figure 3. Which Single Metric Best Predicts Next-Season ERA?

The results were:

Predictor in year (y)	Predictive (R^2) for ERA in (y+1)
SIERA	0.206
FIP	0.199
xFIP	0.195
K-BB%	0.178
K%	0.176
WAR	0.145
ERA	0.116
WHIP	0.101
HR/9	0.059
GB%	-0.009
BB%	-0.009
BABIP	-0.012

SIERA wins.

But only narrowly.

FIP and xFIP are very close behind it.

That is exactly what we might expect from the correlation matrix. If several statistics contain much of the same information, their predictive performance should often be similar.

K-BB% may be the most impressive result in the table.

It is remarkably simple:

\mathrm{K\!-\!BB\%} = K\% - BB\%

Yet its predictive (R²) reaches approximately 0.178.

That is not far behind FIP, xFIP, and SIERA.

BABIP performs particularly poorly. Its cross-validated (R²) is slightly negative.

A negative predictive (R²) does not mean that higher BABIP magically predicts lower ERA.

It means something simpler.

For this particular prediction problem, the fitted BABIP model performs slightly worse than simply predicting the average ERA.

What Happens If We Put Everything Into One Regression?

Here is where things become interesting.

I constructed a full ordinary least-squares regression containing:

ERA, WHIP, FIP, xFIP, SIERA, K%, BB%, HR/9, BABIP, and GB%.

K-BB% was omitted because including K%, BB%, and K-BB% together would introduce exact linear dependency.

The model then produced something strange.

The standardized coefficient for WHIP was approximately:

\beta_{\mathrm{WHIP}} \approx -0.389

Taken literally, that would imply that a higher WHIP predicts a lower future ERA, once the other statistics are held constant.

That is not a sensible baseball interpretation.

It is a multicollinearity problem.

Figure 4. OLS, Ridge, and LASSO Coefficients

Ordinary least squares tries to divide explanatory credit among variables that contain overlapping information.

That can make individual coefficients unstable.

One variable gets a large positive coefficient.

Another highly related variable gets a negative coefficient.

A small change in the sample can alter them again.

The overall model can still predict reasonably well.

The individual coefficients, however, become difficult to interpret.

This is exactly why simply adding every available baseball statistic to a regression is not necessarily a good idea.

More variables do not automatically produce more knowledge.

Sometimes they produce more confusion.

Ridge Regression

Ridge regression offers one solution.

Rather than allowing coefficients to become arbitrarily large, Ridge penalizes them:

\min_{\beta} \left[ \sum_{i=1}^{n} \left( y_i-\widehat{y}_i \right)^2 + \lambda \sum_{j=1}^{p} \beta_j^2 \right]

The tuning parameter (lambda) determines how strongly the coefficients are shrunk toward zero.

When the predictors contain large amounts of overlapping information, this can make the model considerably more stable.

That is exactly what happened.

Under ordinary least squares, the standardized WHIP coefficient was approximately -0.389.

Under Ridge regression, it became approximately:

\beta_{\mathrm{WHIP,Ridge}} \approx 0.013

Essentially zero.

That is an important result.

The model is not saying WHIP is useless.

It is saying:

Once all the other pitching information is already known, WHIP contributes very little additional information about next-season ERA.

That is a very different statement.

LASSO: Let the Model Throw Statistics Away

LASSO takes the regularization idea further.

Instead of penalizing squared coefficients, it penalizes their absolute values:

\min_{\beta} \left[ \sum_{i=1}^{n} \left( y_i-\widehat{y}_i \right)^2 + \lambda \sum_{j=1}^{p} \left| \beta_j \right| \right]

This has an interesting consequence.

LASSO can set some coefficients exactly to zero.

That turns our original question into an empirical experiment.

Give the model all these pitching statistics.

Then ask:

Which ones does it decide it does not need?

In the full-sample fit using the cross-validated penalty, LASSO retained four nonzero predictors:

FIP

SIERA

K%

HR/9

ERA went to zero.

WHIP went to zero.

xFIP went to zero.

BB%, BABIP, and GB% went to zero.

This does not mean those discarded statistics contain no useful baseball information.

It means that, for predicting next-season ERA after the retained variables were already available, their additional contribution was small enough that LASSO discarded them.

That is precisely the kind of redundancy we set out to investigate.

How Stable Was LASSO’s Decision?

A single LASSO fit is useful, but correlated predictors can substitute for one another.

So I also examined how often each statistic survived across the outer cross-validation folds.

Figure 5. Which Metrics Does LASSO Keep?

The approximate selection frequencies were:

Metric	Selected
FIP	100%
SIERA	100%
K%	100%
HR/9	90%
ERA	48%
BB%	24%
xFIP	19%
WHIP	10%
BABIP	10%
GB%	5%

Three statistics survived every time:

FIP, SIERA, and K%.

HR/9 survived in approximately 90 percent of the folds.

WHIP survived only about 10 percent of the time.

That is particularly interesting after the previous WHIP study.

WHIP is very good at describing current run prevention.

Yet once the regression already knows FIP, SIERA, strikeout rate, home-run rate, and the other variables, WHIP rarely contains enough unique predictive information to survive LASSO.

That is not a contradiction.

It is the distinction between useful information and unique information.

Does the Giant Model Actually Predict Better?

Now we arrive at the question that matters most.

Perhaps all this redundancy does not matter if the large model predicts much better.

Does it?

Figure 6. More Metrics Help, but Only a Little

The cross-validated results were:

Model	Predictive (R^2)
ERA only	0.116
SIERA only	0.206
Five-skill model	0.215
All metrics, OLS	0.213
All metrics, Ridge	0.222
All metrics, LASSO	0.217

The five-skill model used K%, BB%, HR/9, BABIP, and GB%.

The best model was the full Ridge regression.

Its predictive (R²) was approximately:

R^2_{\mathrm{Ridge}} = 0.222

SIERA alone produced:

R^2_{\mathrm{SIERA}} = 0.206

The improvement was therefore:

\Delta R^2 = 0.222 - 0.206 = 0.016

Just 1.6 percentage points.

The RMSE tells essentially the same story.

SIERA alone produced an RMSE of about 0.737 ERA runs.

The full Ridge model reduced that to approximately 0.729.

The larger model is better.

But only a little.

That may be one of the most important results in the study.

We gave the model a large collection of modern pitching statistics.

Most of the additional information barely moved the prediction.

Once We Know SIERA, What Else Helps?

This provides another way to look at redundancy.

Start with SIERA, the strongest individual predictor.

Then add other statistics one at a time.

Figure 7. Once We Know SIERA, Most Extra Metrics Add Very Little

SIERA alone:

R^2 = 0.206

Add FIP and the result improves to approximately:

R^2 = 0.217

That is a genuine, although modest, improvement.

Add xFIP to SIERA and predictive performance actually slips slightly, to approximately 0.204.

Add WHIP and it falls to roughly 0.203.

Adding K-BB% changes almost nothing.

Even combining SIERA, FIP, and xFIP reaches only about 0.216.

This is a remarkably clean demonstration of statistical redundancy.

Three statistics are not necessarily three times as informative as one.

Sometimes the second statistic is largely repeating the first.

The third repeats them both.

Why xFIP and SIERA Are a Good Example

Consider again:

r_{\mathrm{xFIP},\mathrm{SIERA}} = 0.970

If two statistics move almost perfectly together, there is simply not much room for one to add entirely new predictive information after the other is already known.

They can still differ conceptually.

Their formulas can still have different purposes.

Their disagreements can even be analytically useful.

But conceptual difference does not guarantee statistical independence.

That distinction is central to this study.

What About WAR?

ERA is only one possible definition of future pitching success.

So I repeated the regression analysis using next-season WAR as the target.

Current WAR itself was the strongest individual predictor:

R^2_{\mathrm{WAR}_y,\mathrm{WAR}_{y+1}} \approx 0.336

SIERA reached approximately 0.252.

FIP reached about 0.243.

xFIP was around 0.235.

WHIP was considerably weaker at approximately 0.125.

Figure 8. Predicting Next-Season WAR

The multivariable models produced:

Model	Predictive (R2)
WAR only	0.336
WAR + IP	0.336
Skill model + IP	0.297
All metrics OLS	0.373
All metrics Ridge	0.375
All metrics LASSO	0.377

Here the large models add somewhat more information.

The best LASSO model improves predictive (R²) from roughly 0.336 using current WAR alone to approximately 0.377.

That is a more noticeable gain than we observed when predicting ERA.

Still, the central lesson remains.

Adding many statistics helps.

But the improvement is nowhere near proportional to the number of variables added.

The Strange WHIP Coefficient Revisited

The negative WHIP coefficient in the ordinary regression is worth returning to because it demonstrates an important statistical point.

WHIP by itself predicts future ERA in the expected direction.

Higher WHIP is associated with higher future ERA.

But once we tell an ordinary regression to hold ERA, FIP, xFIP, SIERA, K%, BB%, HR/9, BABIP, and GB% constant, the WHIP coefficient becomes negative.

Those are two very different questions.

The simple regression asks:

What happens to future ERA when WHIP changes?

The giant multiple regression asks:

What happens when WHIP changes while an enormous collection of closely related pitching statistics somehow remains fixed?

That second scenario may have very little resemblance to an actual pitcher.

The variables are too interconnected.

OLS nevertheless tries to divide the shared information among them.

The resulting coefficient is mathematically legitimate.

Its baseball interpretation is questionable.

Ridge responds by shrinking the WHIP coefficient almost to zero.

LASSO simply removes WHIP.

Both approaches give us a more sensible picture of what the data are saying.

How Many Statistics Do We Really Need?

There is no universal answer.

It depends on the question.

If I want to know how successfully a pitcher kept runners off base, WHIP is excellent.

If I want to know how many earned runs he actually allowed, ERA tells me exactly that.

If I want a simple forward-looking estimate of next-season ERA, SIERA performed best among the individual metrics examined here.

If I want to squeeze out every additional bit of predictive accuracy, a regularized multivariable model performs somewhat better.

But the key word is somewhat.

Going from SIERA alone to a ten-variable Ridge regression improved predictive (R²) from about 0.206 to 0.222.

The sophisticated model wins.

Barely.

The Bigger Lesson

A modern baseball leaderboard can create an illusion of enormous amounts of independent information.

Twenty columns look like twenty facts.

Statistically, that may not be true.

Strikeout ability appears directly in K%.

It appears again in K-BB%.

It also enters FIP, xFIP, and SIERA.

Walks do the same.

Home runs affect FIP and other estimators.

BABIP influences hit-based measures.

ERA and WHIP share the consequences of allowing baserunners.

The columns multiply faster than the underlying baseball phenomena.

That is not a criticism of advanced statistics.

Different metrics were created to answer different questions. They emphasize different aspects of performance. They make different assumptions. They may be useful in different contexts.

The mistake would be assuming that because two statistics have different names, they must contain completely different information.

They often do not.

Conclusion

The original question was simple:

How many pitching statistics do we really need?

The answer is more interesting than I expected.

Many pitching metrics are strongly correlated.

Some are extraordinarily correlated.

xFIP and SIERA correlate at approximately 0.97. ERA and WHIP correlate at about 0.81. FIP, xFIP, SIERA, strikeout measures, walk measures, and home-run measures overlap so heavily that placing all of them in the same ordinary regression produces severe multicollinearity.

The VIF analysis shows the problem.

The unstable OLS coefficients make it visible.

Ridge regression controls it.

LASSO begins throwing redundant statistics away.

And the predictive results show just how little we lose by simplifying.

SIERA alone explains about 20.6 percent of next-season ERA variation under leave-one-season-out prediction.

A full ten-variable Ridge model improves that to approximately 22.2 percent.

Ten statistics are better than one.

But not by much.

That may be the most important conclusion.

The goal should not be to collect the largest possible number of metrics.

It should be to identify statistics that represent genuinely different dimensions of pitching.

After that point, we increasingly begin measuring the same underlying abilities again.

And again.

Just under different names.

 

WHIP may be one of baseball’s most intuitive pitching statistics.

It asks a straightforward question: how many runners does a pitcher allow to reach base by hit or walk per inning pitched?

\mathrm{WHIP}=\frac{BB+H}{IP}

Lower is better.

There is an appealing logic to the statistic. A pitcher who keeps runners off base should allow fewer runs. A pitcher who repeatedly permits hits and walks should eventually pay for them.

And, within a single season, that logic works remarkably well.

But prediction is a different problem.

The fact that WHIP describes a pitcher’s current performance does not necessarily mean that it tells us much about what he will do next year.

So I wanted to test a more difficult question:

How predictive is WHIP?

The Data

I used the season-level FanGraphs data covering 2002 through 2025. The exports include season, innings pitched, ERA, FIP, xFIP, WAR, BABIP, and related measures. A second export supplies the underlying hits and walks needed to examine the components of WHIP. The rate-statistics file contains WHIP itself along with K%, BB%, K-BB%, ERA-, FIP-, xFIP-, FIP, xFIP, and SIERA.

I imposed a minimum of 100 innings:

IP_{i,y}\geq100

That produced 1,699 qualifying pitcher-seasons.

For the year-to-year analysis, a pitcher had to reach 100 innings in consecutive seasons:

IP_{i,y}\geq100\quad\text{and}\quad IP_{i,y+1}\geq100

That left 927 consecutive-season pairs involving 299 different pitchers.

The shortened 2020 season effectively drops out under this rule. I also checked a less restrictive version allowing the following season to fall below 100 innings. The conclusions barely changed, so the principal findings are not an artifact of the cutoff.

First, Does WHIP Describe Current Performance?

Very much so.

The relationship between WHIP and same-season ERA is striking.

Figure 1. Same-Season WHIP and ERA

The fitted relationship is:

\mathrm{ERA}=-1.591+4.316(\mathrm{WHIP})

with:

R^2=0.657

The correlation is approximately:

r=0.811

So WHIP explains about 65.7 percent of the observed variation in same-season ERA among these pitcher-seasons.

That is enormous for such a simple statistic.

This should not be surprising. Runs require baserunners, and WHIP measures two of the principal ways pitchers put those runners on base.

But this is description, not prediction.

The real test begins when we move forward one year.

Does WHIP Repeat Itself?

Before asking whether WHIP predicts ERA, it makes sense to ask an even simpler question.

Does a pitcher’s WHIP this year predict his WHIP next year?

Figure 2. Year-to-Year WHIP

The regression is:

\mathrm{WHIP}_{y+1}=0.626+0.511(\mathrm{WHIP}_{y})

with:

R^2=0.224

The year-to-year correlation is approximately 0.474.

That is meaningful.

Pitchers with low WHIPs tend to have relatively low WHIPs again. Pitchers with high WHIPs tend to remain higher.

But look at what happened to the explanatory power.

Same season:

R²=0.657

Next season:

R²=0.224

More than two-thirds of WHIP’s apparent explanatory power disappears when we move the target forward one season.

That is our first indication of regression toward the mean.

The Main Question: Does WHIP Predict Next-Season ERA?

Now we come to the central test.

Figure 3. WHIP and Next-Season ERA

The fitted equation is:

\mathrm{ERA}_{y+1}=1.482+1.918(\mathrm{WHIP}_{y})

with:

R^2=0.110

The correlation is approximately 0.332.

So current WHIP explains only about 11 percent of the variation in next-season ERA.

That is not nothing.

WHIP contains a real signal.

But compare 11 percent with the 66 percent relationship between WHIP and ERA in the same season.

That difference is the heart of the study.

WHIP is much better at telling us what has happened than telling us what will happen.

How Does WHIP Compare With Other Predictors?

This is where the result becomes more interesting.

I compared several current-season statistics as predictors of next-season ERA:

Current-season statistic	Leave-one-season-out predictive (R2)
SIERA	0.206
FIP	0.199
xFIP	0.195
K-BB%	0.178
ERA	0.116
WHIP	0.101
BABIP	-0.012

For this comparison, I used leave-one-season-out validation. Each season was withheld in turn, the regression was estimated using the other seasons, and predictions were generated for the season that had not been used to fit the model.

That makes this a considerably tougher test than simply reporting an in-sample correlation.

Figure 4. Which Statistic Predicts Next-Season ERA Best?

SIERA wins.

FIP and xFIP are essentially right behind it.

K-BB% performs surprisingly well considering how simple it is.

Current ERA beats WHIP, although not by much.

And BABIP is essentially useless for predicting next-season ERA in this sample. Its cross-validated (R^2) is actually slightly below zero, meaning that simply predicting the overall mean would perform marginally better.

This ranking makes baseball sense.

FIP, xFIP, and SIERA deliberately concentrate more heavily on pitcher-controlled characteristics such as strikeouts, walks, and home runs.

WHIP includes hits.

That is both its strength and its weakness.

Hits matter enormously right now.

They are less stable going forward.

Regression Toward the Mean

One of the clearest ways to see the problem is to divide the pitchers into five groups based on their WHIP in year (y).

Figure 5. WHIP Quintiles and Regression Toward the Mean

The best-WHIP quintile averaged approximately:

1.054 WHIP and 2.97 ERA in the current season.

Their average ERA the following year?

3.47.

That is still very good, but nowhere near 2.97.

At the other extreme, the worst-WHIP quintile averaged approximately:

1.454 WHIP and 4.69 ERA.

The following season their ERA improved to approximately:

4.22.

The extremes move toward one another.

The spread between the best and worst groups was about 1.72 ERA runs in year (y).

One season later, it was only about 0.76 runs.

WHIP clearly contains information that persists.

It simply does not persist nearly as strongly as the current-season relationship might lead us to believe.

Which Part of WHIP Is Responsible?

WHIP consists of two pieces:

\mathrm{WHIP}=\frac{H}{IP}+\frac{BB}{IP}

That gives us a useful experiment.

We can separate hits from walks.

Using per-nine-inning versions:

H/9=9\left(\frac{H}{IP}\right)

and:

BB/9=9\left(\frac{BB}{IP}\right)

The results contain a small surprise.

Figure 6. What Part of WHIP Persists?

BB/9 is actually the most stable individual component from one year to the next.

Its year-to-year (R²) is approximately:

0.425

H/9 has a year-to-year (R²) of about:

0.257

and WHIP itself:

0.224

So walk rate is considerably more persistent than hit rate.

But when the target is next-season ERA, the story reverses.

Current H/9 explains about 10.4 percent of future ERA variation.

Current BB/9 explains only about 1.1 percent.

WHIP explains about 11.0 percent.

That sounds contradictory at first.

It is not.

Walk rate is a repeatable pitcher skill, but within this group of durable pitchers its variation alone does not explain very much of next year’s ERA. Hits allowed are less stable, yet they contain information related to strikeout ability, contact suppression, and other pitcher characteristics.

Importantly, BABIP itself had essentially no relationship with next-season ERA.

So the useful information in H/9 is not simply “this pitcher had a low BABIP, therefore he will again.”

Does WHIP Add Anything to Better Pitching Metrics?

This may be the most revealing test of all.

Suppose we already know a pitcher’s FIP or SIERA.

Does knowing his WHIP improve our prediction?

Figure 7. Does WHIP Add Predictive Information?

The answer is essentially no.

The leave-one-season-out results were approximately:

Model	Predictive (R2)
ERA	0.116
ERA + WHIP	0.119
FIP	0.199
FIP + WHIP	0.197
SIERA	0.206
SIERA + WHIP	0.203

Adding WHIP to ERA produces a tiny improvement.

Adding WHIP to FIP actually lowers out-of-sample performance slightly.

Adding it to SIERA does the same.

That does not mean WHIP is worthless.

It means that once we know a more sophisticated pitcher estimator, WHIP contributes little additional predictive information.

SIERA already knows much of what we need to know about the pitcher’s underlying skills.

WHIP mainly tells us what happened to those skills and batted balls in the season we just observed.

A League-Adjusted Check

There is another potential concern.

Baseball’s run environment changed between 2002 and 2025.

Fortunately, the supplied FanGraphs data also contain ERA-, FIP-, xFIP-, and league-adjusted WHIP+.

Repeating the year-to-year analysis with those adjusted measures does not change the conclusion.

Current WHIP+ explains about 9.4 percent of next-season ERA- variation.

Current ERA- explains about 10.1 percent.

FIP- rises to roughly 17.6 percent, while xFIP- reaches approximately 17.9 percent.

The same hierarchy remains.

The finding is therefore not merely the result of changing league scoring environments.

What WHIP Is Good For

None of this makes WHIP a bad statistic.

Quite the opposite.

A statistic explaining roughly 66 percent of the variation in same-season ERA with a formula consisting only of hits, walks, and innings is extraordinarily efficient.

WHIP answers a very useful question:

How successfully has this pitcher kept runners off base?

It is easy to calculate.

Easy to understand.

Easy to compare.

And strongly connected to what happened on the scoreboard.

But a descriptive statistic and a predictive statistic are not the same thing.

That distinction matters.

FIP, xFIP, SIERA, and K-BB% sacrifice some of WHIP’s intuitive simplicity in an attempt to isolate skills that are more likely to persist.

Our results suggest that tradeoff works.

Conclusion

WHIP passes one test spectacularly and another only modestly.

As a description of current performance, it is excellent.

R^2_{\mathrm{WHIP,\ same\ season\ ERA}}=0.657

As a predictor of its own next-season value:

R^2_{\mathrm{WHIP}_y,\mathrm{WHIP}_{y+1}}=0.224

And as a predictor of next-season ERA:

R^2_{\mathrm{WHIP}_y,\mathrm{ERA}_{y+1}}=0.110

Under the stricter leave-one-season-out test, that last value falls slightly to about 0.101.

SIERA roughly doubles that predictive performance.

FIP and xFIP come close.

K-BB% also comfortably beats WHIP.

So the answer to the original question is fairly clear.

WHIP is predictive, but not especially predictive.

A low WHIP today is meaningful evidence that a pitcher will perform well tomorrow. It is simply much weaker evidence than its strong relationship with today’s ERA might suggest.

That may be WHIP’s most interesting characteristic.

It looks like a forward-looking statistic because it is so closely connected with pitching success.

In reality, much of its power belongs to the present.

WHIP tells us exceptionally well what a pitcher has done.

For predicting what he will do next, we can do better.

 

Introduction

This project began with a simple question:

Who were the greatest third basemen?

But the longer the study went on, the less simple that question became.

Third base is not first base. It is not a position where offense alone usually defines the job. But it is also not shortstop or second base, where defensive value and up-the-middle responsibility have historically borne the greater burden.

Third base lives between those worlds.

It is a reaction position.
It is an arm-strength position.
It is a power position.
It is a two-way position.

That is why third base is so interesting analytically. The position resists one-dimensional ranking. A player can dominate the position through offense, as Mike Schmidt, Eddie Mathews, Chipper Jones, George Brett, and Jose Ramirez did. A player can dominate it through defense, as Brooks Robinson did. A player can become historically important through balance, as Scott Rolen, Adrian Beltre, Nolan Arenado, and Wade Boggs did in different ways.

The point of this final chapter is not to introduce another ranking. It is to bring the project together.

The central claim is this:

Third base is baseball’s hybrid position.

It is not merely a corner power position.
It is not merely a defensive position.
It is the place where power, reaction defense, arm strength, durability, and era context all meet.

A Note on Era Context

Comparing baseball players across eras is difficult. The game has changed. The league has expanded. The talent pool has changed. Integration, globalization, training, travel, medicine, equipment, ballparks, strategy, and offensive environments all matter.

The Full House Modeling paper by Yan, Burgos, Kinson, and Eck makes this point directly. Their framework builds on Stephen Jay Gould’s “full house” idea by treating player achievement as part of a broader distribution of performance, while also considering the changing baseball talent pool over time. Their method is more ambitious than the one used here because it explicitly models talent-pool size and era-adjusted performance.

This project does not implement Full House Modeling. It uses a narrower, more transparent adjustment:

Compare third basemen to other third basemen in the same season.

That is the role of the z-score method.

z_{i,y,m} = \frac{ x_{i,y,m} - \overline{x}_{p,y,m} }{ s_{p,y,m} }

Where:

i = \text{player} y = \text{season} p = \text{primary position} m = \text{metric}

In plain English, the question becomes:

How far did this third baseman stand above or below other third basemen in the same season?

That does not solve every era problem. But it does control for one of the most important things: the player’s immediate positional environment.

Defining the Position

Before ranking third basemen, we first need to understand the position itself.

For the home run study, each player-season was assigned a primary position based on where the player appeared most often:

\mathrm{PrimaryPosition}_{i,y} = \operatorname*{arg\,max}_{p \in \mathcal{P}} G_{i,y,p}

Where:

G_{i,y,p} = \text{games played by player } i \text{ at position } p \text{ in season } y

The main home run study used regular player-seasons:

PA \geq 300

The full time frame was:

1871–2025

That gave us a broad historical view of how home runs have been distributed by position.

Figure 1: Total Home Runs by Position

Figure 1. Total home runs by primary position, regular player-seasons, PA ≥ 300, 1871–2025.

The total home run bar chart gives the project a useful starting point. It shows where the historical home run volume has come from.

For regular player-seasons:

\mathrm{TotalHR}_{1B} = 48{,}064 \mathrm{TotalHR}_{RF} = 42{,}977 \mathrm{TotalHR}_{LF} = 40{,}663 \mathrm{TotalHR}_{3B} = 36{,}127

Third base ranks behind first base and the corner outfield positions, but ahead of center field, catcher, second base, and shortstop.

That matters.

It shows that third base has historically been a significant power source. It is not a middle-infield power-light position. But it is not first base either. Third base carries a heavier defensive burden than first base and a higher power expectation than second base or shortstop.

That is the hybrid identity of the position.

The total home run formula is simple:

\mathrm{TotalHR}_{p} = \sum_{i,y} HR_{i,y,p}

The conclusion is not subtle:

Third base belongs among the power-producing positions,

but it is not purely a power position.

Figure 2: Home Run Distribution by Position

Figure 2. Home runs by primary position, regular player-seasons, PA ≥ 300, 1871–2025.

The box plots add another layer.

The bar chart showed total historical volume. The box plot shows the distribution of individual player-seasons.

For regular player-seasons, the median home runs by position were:

DH: 18

1B: 12

RF: 11

LF: 10

3B: 9

C: 7

CF: 7

2B: 5

SS: 4

In compact form:

\widetilde{\mathrm{HR}}_{\mathrm{DH}} > \widetilde{\mathrm{HR}}_{\mathrm{1B}} > \widetilde{\mathrm{HR}}_{\mathrm{RF}} > \widetilde{\mathrm{HR}}_{\mathrm{LF}} > \widetilde{\mathrm{HR}}_{\mathrm{3B}} > \widetilde{\mathrm{HR}}_{\mathrm{C}} \approx \widetilde{\mathrm{HR}}_{\mathrm{CF}} > \widetilde{\mathrm{HR}}_{\mathrm{2B}} > \widetilde{\mathrm{HR}}_{\mathrm{SS}}

Again, third base sits in the middle-to-upper part of the power spectrum.

That position explains why third-base rankings are hard. A third baseman who hits like a first baseman is unusually valuable. A third baseman who fields like an elite shortstop-adjacent defender is also unusually valuable. A third baseman who does both is historically rare.

The Offensive Summit

The offensive studies showed that Mike Schmidt remains the defining offensive third baseman.

Using the Model C offensive framework, the season score combined on-base ability, isolated power, walk rate, strikeout avoidance, baserunning, run scoring, and run production.

\begin{aligned} \mathrm{ModelCOffensiveScore} &= z_{\mathrm{OBP}} + z_{\mathrm{ISO}} + z_{\mathrm{BB/PA}} + z_{\mathrm{LowSO/PA}} \\ &\quad+ z_{\mathrm{NetSB/PA}} + z_{\mathrm{R/PA}} + z_{\mathrm{RBI/PA}} \end{aligned}

The career score was the sum of weighted season scores:

\mathrm{CareerScore} = \sum_s \mathrm{WeightedSeasonScore}_{s}

And peak value was measured through a player’s best seven seasons:

\mathrm{Peak7Score} = \sum_{k=1}^{7} \mathrm{BestSeasonScore}_{k}

Schmidt’s greatness came from both career value and peak value. He was not merely a compiler. He was not merely a short-peak slugger. He combined long-term production with enormous seasonal dominance.

The offensive group around him included Eddie Mathews, Chipper Jones, George Brett, Jose Ramirez, Alex Rodriguez, Wade Boggs, Home Run Baker, Ron Santo, and David Wright.

But they were not all the same kind of offensive player.

Schmidt was power and patience.
Boggs was on-base dominance.
Brett was batting skill and all-around production.
Chipper was switch-hitting offensive force.
Jose Ramirez was compact modern power, baserunning, and peak efficiency.
Mathews was long-career left-handed power.

The offensive rankings showed that third-base greatness has several offensive forms.

The Defensive Summit

The traditional defensive study told a different story.

The defensive model used assists per game, putouts per game, double plays per game, fielding percentage, and low errors per game.

\mathrm{TraditionalDefensiveScore} = z_{\mathrm{A/G}} + z_{\mathrm{PO/G}} + z_{\mathrm{DP/G}} + z_{\mathrm{FPct}} + z_{\mathrm{LowE/G}}

The defensive peak and career results were clear:

Brooks Robinson is the traditional defensive summit.

Robinson stood apart because his defensive career was both excellent and enormous. He was not only a peak defender. He was a long-duration defensive institution.

The next group included Nolan Arenado, Gary Gaetti, Willie Kamm, Buddy Bell, Clete Boyer, Adrian Beltre, Scott Rolen, Matt Chapman, Graig Nettles, Mike Lowell, and Wade Boggs.

This defensive list is important because it changes the shape of the third-base conversation.

If we looked only at offense, Brooks Robinson would not appear near the top. If we looked only at defense, Chipper Jones would not appear near the top. Third base requires us to keep both dimensions visible.

Figure 3: Offense and Defense Together

Figure 3. Career offensive score versus traditional defensive score for third basemen.

The two-dimensional scatterplot is one of the most important figures in the project.

It shows that third basemen do not fall along a single line from bad to great. They occupy different regions of the offensive-defensive space.

The combined two-way score was built by standardizing the career offensive and defensive dimensions:

\mathrm{CombinedZ}_{i} = z_{\mathrm{Offense},i} + z_{\mathrm{Defense},i}

The top of the combined list included:

Mike Schmidt

Brooks Robinson

Scott Rolen

Nolan Arenado

Wade Boggs

Eddie Mathews

Chipper Jones

Willie Kamm

George Brett

Adrian Beltre

This is the key result of the project.

Mike Schmidt wins because he was historically great on offense and still positive on defense.

Brooks Robinson ranks so high because his defensive value was so extreme that it overcame a more modest offensive profile.

Scott Rolen and Adrian Beltre matter because they represent the two-way ideal. They were not Schmidt-level offensive forces or Brooks-level defensive singularities, but they were excellent in both directions.

Nolan Arenado appears as a modern defense-heavy two-way star. Wade Boggs appears as a batting-and-defense star whose value is not captured by power alone.

Third base becomes clearest when seen as a two-axis position.

Figure 4: Dendrogram of Top Combined Third Basemen

Figure 4. Dendrogram showing similarity among the top 15 third basemen by combined offense-defense score.

The combined top list is not the same as the top offensive or top defensive lists.

That is the point.

A one-stat answer hides the different paths to value. The combined score helps reveal those paths.

The general model is:

\mathrm{TwoWayScore}_{i} = z_{\mathrm{Offense},i} + z_{\mathrm{Defense},i}

But the interpretation matters more than the arithmetic.

Schmidt is the best answer because he dominates the total landscape.
Robinson is the best defensive answer.
Chipper is one of the best offensive answers.
Rolen and Beltre are among the cleanest two-way answers.
Boggs and Brett show that contact, on-base skill, and complete offensive value can compete with pure home run power.
Arenado shows how a modern elite defender can push into the historical top tier.

The final ranking should therefore not be read as a single flat list. It should be read as a map of greatness.

The Center of the Position

The middle of the project may have been just as revealing as the top.

When we searched for the most average third basemen, we found different kinds of centers.

Offensively, Casey Blake represented a kind of Model C center.
Defensively, Eddie Mathews appeared surprisingly close to the traditional defensive center.
In the two-dimensional offense-defense space, Bill Melton emerged as the closest player to the combined center.

The two-dimensional typicality calculation was:

\mathrm{Typicality}_{i} = \sqrt{ z_{\mathrm{Offense},i}^{2} + z_{\mathrm{Defense},i}^{2} }

A low value means the player is close to the center of the third-base population.

This was useful because the average player gives meaning to the extremes.

Without the center, Schmidt is just a big number.
With the center, Schmidt becomes distance from normal.
Without the center, Brooks Robinson is just a defensive legend.
With the center, we can see how unusual his profile really was.

The middle defines the scale.

Figure 5: Closest Third Basemen to the Offense-Defense Center

Figure 5. Third basemen closest to the two-dimensional offense-defense center, measured by typicality.

The Lower Tail

The lower-tail studies also mattered.

Ken Reitz appeared as the lower offensive tail among third-base regulars, but his traditional defensive profile was positive. That makes him more interesting than a simple “worst hitter” label.

Chris Johnson emerged as a lower-tier tailback because he combined a weak offensive standing with a weak traditional defensive standing.

Figure 6: Lowest Combined Offense-Defense Third Basemen

Figure 6. Lowest combined offense-defense scores among third-base regulars.

The lower-tail combined score was:

\mathrm{LowerTailScore}_{i} = z_{\mathrm{Offense},i} + z_{\mathrm{Defense},i}

The lowest combined players included:

Chris Johnson

Butch Hobson

Wes Helms

Larry Parrish

Harry Lord

Tom Brookens

Maikel Franco

Dean Palmer

Charley Smith

Ray Jablonski

The lower tail reinforced a point that runs through the whole project:

A bad offensive third baseman is not necessarily a bad third baseman.

A bad defensive third baseman is not necessarily a bad third baseman.

But a player weak in both dimensions falls quickly.

Third base gives players several ways to survive. But it also exposes players who cannot contribute either way.

Validation with WAR

The WAR validation study gave the project an external check.

The model was not designed to reproduce WAR exactly. But if the offensive and defensive scores were meaningful, they should explain a substantial portion of WAR variation.

At the career level for regular third basemen, the offense-plus-defense model was strong:

\mathrm{WAR} = 14.90 + 0.60(\mathrm{CareerOffensiveScore}) + 0.59(\mathrm{CareerDefensiveScore}) R^2 = 0.814

That is an important result.

It means the simple, transparent framework captures much of what WAR captures, even though WAR is built on a more complex run-value framework.

The validation did not prove the model was perfect. It showed that the model was sensible.

Offense mattered.
Defense mattered.
Together they explained much more than either alone.

Validation with wRC+

The wRC+ validation was equally useful because it directly tested the offensive side.

At the season level, Model C offensive score predicted wRC+ well:

wRC^+ = 101.47 + 5.86(\mathrm{ModelCOffensiveScore}) R^2 = 0.692

At the career level, the offensive score also connected strongly to career wRC+:

wRC^+ = 100.89 + 5.41(\mathrm{AverageOffensiveScore}) R^2 = 0.740

The defensive negative control was also important. Traditional defensive score did not meaningfully predict wRC+. That is what we wanted to see.

The offensive model tracked offense.
The defensive model tracked defense.
The two-way model tracked overall value.

That separation gives the project credibility.

Figure 7: WAR Validation

Figure 7. Career WAR, actual versus predicted from offensive and defensive scores.

The WAR validation figure belongs in the final chapter because it shows that the project is not only descriptive. It is also diagnostic.

The model found recognizable third-base greatness because its components corresponded to real baseball value.

Schmidt, Robinson, Beltre, Rolen, Boggs, Mathews, Brett, Chipper, Arenado, Santo, and others were not artifacts of the scoring system. They appeared because the scoring system captured meaningful dimensions of the position.

Figure 8: wRC+ Validation

Figure 8. Model C offensive score versus wRC+ for third-base seasons.

The wRC+ validation figure should follow the WAR figure.

Together, these two validation figures show the difference between offensive value and total value.

wRC+ confirms the offensive model.
WAR confirms the broader two-way model.

That distinction is exactly what third base requires.

The Dendrogram: Types of Greatness

The dendrogram in Figure 4 adds one final interpretive layer.

Instead of asking who ranked first, it asked which great players were similar.

The distance formula was:

d(i,j) = \sqrt{ \left( z_{\mathrm{Offense},i} - z_{\mathrm{Offense},j} \right)^2 + \left( z_{\mathrm{Defense},i} - z_{\mathrm{Defense},j} \right)^2 }

This produced recognizable clusters.

Schmidt, Chipper, Mathews, Brett, and Jose Ramirez formed an offense-heavy region.

Brooks Robinson stood apart as the defensive summit.

Rolen, Beltre, Arenado, Boggs, Nettles, Kamm, and similar players occupied the two-way or defense-strong region.

The dendrogram reinforces the project’s central point:

There is no single shape of third-base greatness.

There are families of greatness.

As Figure 4 shows, the dendrogram turns the ranking into a typology.

It lets the reader see why direct comparisons can be difficult.

Is Chipper Jones greater than Brooks Robinson?
It depends on what kind of greatness we are asking about.

Is Scott Rolen closer to Adrian Beltre than to Mike Schmidt?
In a two-dimensional offensive-defensive space, yes.

Is Mike Schmidt the most complete answer?
The evidence says yes.

Final Interpretation

After all of the rankings, models, box plots, bar charts, validations, residuals, and dendrograms, the final interpretation is this:

Mike Schmidt is the best overall third baseman in this framework.

Brooks Robinson is the defensive summit.

Chipper Jones is one of the offensive summits.

Scott Rolen and Adrian Beltre are two-way ideals.

Nolan Arenado is the modern defensive-power bridge.

Wade Boggs and George Brett show that third-base offense is not only home runs.

But the broader conclusion is about the position itself.

Third base is a test of balance.

A great third baseman can win with the bat.
A great third baseman can win with the glove.
The greatest third basemen usually do both.

The home run study showed that third base belongs in the power conversation. The defensive study showed that the position still carries serious fielding responsibility. The WAR and wRC+ validations showed that the model behaves sensibly. The center and lower-tail studies showed that the full distribution matters, not only the top.

That brings the project back to the “full house” idea. Baseball greatness is not only about the record holder. It is about where that record, season, or career sits within the full distribution of the game.

This project used third base as the test case.

And third base rewarded the approach.

Conclusion

Third base is baseball’s hybrid position.

It is close enough to first base and the corner outfield to demand power.
It is close enough to shortstop and second base to demand defense.
It is far enough from both extremes to create unusual player types.

That is why the position has produced so many different kinds of great players.

Mike Schmidt towers because he solved the hybrid problem better than anyone else in this framework. He hit like a historic slugger while remaining a positive defensive player at a demanding position.

Brooks Robinson towers because his defense was so exceptional that it reshaped the positional landscape.

Rolen, Beltre, Arenado, Boggs, Brett, Chipper, Mathews, Santo, and others fill out the map because each represents a different solution to the same positional problem.

The final answer, then, is not only a ranking.

It is a definition.

Third base is where power and defense meet.

And that is why it was worth studying.

 

 

Baseball has changed enormously since the middle of the twentieth century.

Strikeouts have increased. Home-run rates have risen and fallen. Stolen bases have moved in and out of fashion. Relief pitching has become more specialized. Defensive positions have acquired different offensive expectations.

It is tempting, therefore, to divide baseball history into a sequence of distinct player types. The contact hitter belongs to one era. The base stealer belongs to another. The modern game belongs to the power hitter who walks frequently and strikes out even more frequently.

But is that really what happened?

Did one type of hitter replace another, or did the same basic offensive archetypes persist across changing statistical environments?

To explore that question, I used the Lahman baseball database to examine 9,218 qualified non-pitcher seasons from 1954 through 2025. I standardized each player against other qualified hitters in his own season, reduced the statistical profiles using principal component analysis, and then used clustering to identify recurring offensive archetypes.

The results reveal six recognizable types of offensive players:

1. Patient-contact hitters
2. Low-impact contact hitters
3. Elite power-and-patience hitters
4. Power-and-strikeout hitters
5. Aggressive free-swingers
6. Speed-and-contact hitters

Perhaps most importantly, the results suggest that baseball’s statistical environment has changed much more dramatically than its underlying distribution of player types.

The modern hitter may look different in the raw statistics. Relative to his contemporaries, however, he often occupies a role that has existed for generations.

Building the Historical Sample

I began with five Lahman tables:

• Batting
• People
• Appearances
• Fielding
• Teams

Players who appeared for multiple teams during one season were combined into a single player-season record. I used the appearances data to assign each player a primary defensive position, defined as the position at which he appeared most frequently during that season.

Pitchers were excluded.

To account for seasons of different lengths, I defined a qualified season using the familiar standard of 3.1 plate appearances per scheduled team game.

Because teams occasionally played slightly different numbers of games, I used the median number of team games during each season:

PA_{i,y} \geq 3.1 \widetilde{G}_{y}

where:

\widetilde{G}_{y} = \operatorname{median}\left(G_y\right)

Here, (\widetilde{G}_{y}) represents the median number of team games played in season (y).

This method adjusts the qualification threshold for 154-game seasons, 162-game seasons, strike-shortened seasons, and the 60-game 2020 season.

The study begins in 1954 because the variables needed for the full model are not consistently complete before that date. Strikeouts become sufficiently complete before then, but caught stealing and sacrifice flies create additional limitations. Beginning in 1954 allows the same five-variable model to be used across the entire study period.

Figure 1. The number of qualified player-seasons generally increased as Major League Baseball expanded.

The increase in qualified seasons primarily reflects league expansion. The early portion of the study contains fewer than 100 qualified hitters in many seasons. By the late 1990s and early 2000s, the total frequently exceeded 150.

Measuring an Offensive Profile

I wanted to describe how a hitter produced offense, not merely how much offense he produced.

I therefore selected five variables:

• On-base percentage
• Isolated power
• Walks per plate appearance
• Strikeouts per plate appearance
• Net stolen bases per plate appearance

Plate appearances were calculated as:

PA = AB + BB + HBP + SF + SH

On-base percentage was calculated as:

\mathrm{OBP} = \frac{ H + BB + HBP }{ AB + BB + HBP + SF }

Isolated power measures extra-base power beyond batting average:

\mathrm{ISO} = \mathrm{SLG} - \mathrm{AVG}

The walk and strikeout rates were:

\mathrm{BB/PA} = \frac{BB}{PA} \mathrm{SO/PA} = \frac{SO}{PA}

Finally, I defined net stolen-base production as:

\mathrm{NetSB/PA} = \frac{ SB - CS }{ PA }

This final measure rewards successful steals while penalizing caught-stealing events.

I did not include runs or RBI in the clustering model. Both statistics are strongly affected by batting order, teammates, and opportunity. They tell us something about the results of a player’s season, but less about the underlying style with which he produced those results.

Home runs were also not included as a separate rate because ISO already captures power production. Adding both ISO and home runs per plate appearance would have given power disproportionate weight in the clustering.

Baseball’s Changing Offensive Environment

The raw statistics immediately demonstrate how much the offensive environment changed.

Figure 2. Mean offensive rates among qualified hitters, 1954-2025.

In 1954, the average qualified hitter in the sample had approximately:

\mathrm{OBP} = 0.355 \mathrm{ISO} = 0.151 \mathrm{SO/PA} = 0.088

By 2025, the corresponding values were approximately:

\mathrm{OBP} = 0.330 \mathrm{ISO} = 0.178 \mathrm{SO/PA} = 0.204

The average qualified hitter’s strikeout rate more than doubled.

Power generally increased, particularly during the offensive surge of the late 1990s and again during the home-run-heavy seasons of the late 2010s. Walk rates changed much less. On-base percentage remained within a fairly narrow historical range, although it rose noticeably around 2000 before declining.

Stolen-base production followed a different pattern. It increased during the 1970s and 1980s, declined during the power-oriented environment that followed, and has recently begun to rise again.

These changes create a serious comparison problem. A 20 percent strikeout rate would have been extraordinary during much of the twentieth century. In the modern game, it may be close to ordinary.

A player cannot be classified historically based solely on raw statistics.

Adjusting Every Player for His Era

To make the seasons comparable, I standardized each variable within its own season.

For player (i), season (y), and offensive measure (m):

z_{i,y,m} = \frac{ x_{i,y,m} - \overline{x}_{y,m} }{ s_{y,m} }

A value of: z=0 indicates that the player was equal to the seasonal average.

A value of: z=1 indicates that he was one standard deviation above the seasonal average.

This adjustment alters the analysis’s meaning. I am not asking whether a hitter had a high strikeout rate in absolute terms. I am asking whether he struck out frequently compared with the other qualified hitters of his own season.

A player from 1965 and a player from 2025 can therefore belong to the same archetype even though their raw statistics differ substantially.

They occupied the same relative position within their respective baseball environments.

Reducing the Offensive Dimensions

The five standardized variables remain related to one another. High-OBP players often walk frequently. Power hitters may also strike out frequently. Speed-oriented players tend to have different power and contact profiles.

I used principal-component analysis to summarize these relationships.

The first principal component explained: 42.5% of the total variation.

The second explained: 25.0%. Together, the first two components explained: 42.5% +25.0% = 67.5%

The first three components explained approximately: 86.2% of the total variation.

The first component primarily represents overall offensive force. OBP, ISO, and walk rate all load positively on this dimension. Players far to the right of the PCA plot tend to reach base, hit for power, and draw walks at rates well above their seasonal environments.

The second component separates high-power, high-strikeout hitters from lower-strikeout players with stronger contact or speed characteristics.

How Many Archetypes Are There?

Clustering always requires a choice about how much detail to preserve.

A model with too few clusters combines meaningfully different players. A model with too many clusters produces distinctions that may be statistically fragile or difficult to interpret.

The k-means procedure attempts to minimize the total squared distance between each player and the center of his assigned cluster:

\mathrm{WCSS} = \sum_{k=1}^{K} \sum_{i \in C_k} \left\lVert \mathbf{z}_i - \boldsymbol{\mu}_k \right\rVert^2

where:

C_k=\mathrm{cluster}\ k

\mathbf{z}_i = \mathrm{standardized\ profile\ of\ player\!-\!season}\ i

\boldsymbol{\mu}_{k} = \mathrm{center\ of\ cluster}\ k

I tested solutions containing four through nine clusters.

Figure 3. The silhouette score is highest for the four-cluster model, although the six-cluster model retains useful baseball distinctions.

The silhouette statistic compares the average distance between a player and his own cluster with the distance between that player and the nearest alternative cluster:

s(i) = \frac{ b(i)-a(i) }{ \max\left\{a(i),b(i)\right\} }

The four-cluster solution produced the strongest formal separation. However, it merged several historically meaningful offensive styles.

In particular, it tended to combine elite power-and-patience hitters with less complete power hitters, and it reduced distinctions between contact-oriented players.

I therefore selected the six-cluster solution.

This is an interpretive decision. The six clusters are not six perfectly isolated biological species. Baseball players exist along continuous statistical dimensions. The purpose of the clusters is to provide a useful map of that continuum.

Figure 4. The 9,218 player-seasons displayed in the space formed by the first two principal components.

The overlap in Figure 4 is important. The clusters are recognizable, but their boundaries are not absolute. A player near a boundary may resemble members of two neighboring archetypes.

The Six Offensive Archetypes

Figure 5 shows the standardized center of each cluster.

Figure 5. Each value represents the number of standard deviations above or below the seasonal mean.

1. Patient-Contact Hitters

Player-seasons: 2,190
Share of sample: 23.8 percent

The patient-contact group combines:

• Above-average OBP
• Above-average walk rates
• Low strikeout rates
• Approximately average power
• Below-average stolen-base production

Its cluster center has an OBP score of:

z_{\mathrm{OBP}} = +0.55

and a strikeout-rate score of:

z_{\mathrm{SO/PA}} = -0.50

These hitters generally controlled the strike zone and put the ball in play. They were not necessarily powerless, but power was not the defining feature of the group.

The most statistically central examples include:

• Justin Turner, 2022
• Eric Hosmer, 2015
• Edgardo Alfonzo, 1999

These are not necessarily the greatest seasons in the cluster. They are the seasons located closest to its statistical center.

2. Low-Impact Contact Hitters

Player-seasons: 1,925
Share of sample: 20.9 percent

This group also struck out infrequently, but without the OBP, walks, or power of the patient-contact cluster.

Its center was:

z_{\mathrm{OBP}} = -0.74 z_{\mathrm{ISO}} = -0.94 z_{\mathrm{BB/PA}} = -0.81 z_{\mathrm{SO/PA}} = -0.84

These hitters made contact, but much of that contact produced limited offensive value. Their low strikeout rates should not automatically be interpreted as evidence of superior hitting.

This distinction is important. Avoiding strikeouts is valuable only when the resulting balls in play produce enough hits, power, or advancement to compensate for the lost walks and extra-base production.

Central examples include:

• Danny Bautista, 2004
• Marlon Anderson, 2002
• Melky Cabrera, 2007

3. Elite Power-and-Patience Hitters

Player-seasons: 851
Share of sample: 9.2 percent

This is the smallest cluster and the most offensively dominant.

Its center was approximately:

z_{\mathrm{OBP}} = +1.74 z_{\mathrm{ISO}} = +1.25 z_{\mathrm{BB/PA}} = +1.74

The cluster’s strikeout rate was almost exactly average relative to each season:

z_{\mathrm{SO/PA}} = +0.05

These hitters combined elite on-base ability with elite power and patience. Unlike the power-and-strikeout group, they did not require an exceptionally high strikeout rate to produce their power.

Central examples include:

• Al Kaline, 1966
• Kris Bryant, 2017
• Ben Zobrist, 2009

The presence of players from widely separated eras is exactly what the season adjustment was designed to reveal. Their raw strikeout totals and league environments differed, but their relative offensive structures were similar.

4. Power-and-Strikeout Hitters

Player-seasons: 1,363
Share of sample: 14.8 percent

This group most closely resembles the familiar three-true-outcomes hitter.

Its defining characteristics were:

z_{\mathrm{ISO}} = +1.00 z_{\mathrm{BB/PA}} = +0.62 z_{\mathrm{SO/PA}} = +1.25

The group produced power and drew walks, but also struck out much more frequently than its seasonal peers.

Its OBP remained modestly above average:

z_{\mathrm{OBP}} = +0.25

Central examples include:

• Jack Clark, 1979
• Andruw Jones, 2002
• Dale Murphy, 1986

This cluster existed long before the recent explosion in league-wide strikeouts. The modern environment made the raw statistical profile more common, but the relative archetype was already present.

5. Aggressive Free-Swingers

Player-seasons: 1,946
Share of sample: 21.1 percent

The aggressive free-swinging group had:

• Below-average OBP
• Below-average walk rates
• Above-average strikeout rates
• Approximately average power
• Little baserunning contribution

Its center included:

z_{\mathrm{OBP}} = -0.83 z_{\mathrm{BB/PA}} = -0.66 z_{\mathrm{SO/PA}} = +0.66

Power was only slightly above the seasonal average:

z_{\mathrm{ISO}} = +0.06

This is an important contrast with the power-and-strikeout cluster. Both groups struck out frequently, but the aggressive free-swingers did not receive the same compensating power, walks, or OBP.

Central examples include:

• Ollie Brown, 1969
• Ryan Ludwick, 2009
• Matt Williams, 1998

6. Speed-and-Contact Hitters

Player-seasons: 943
Share of sample: 10.2 percent

This was the most specialized cluster.

Its net stolen-base score was:

z_{\mathrm{NetSB/PA}} = +2.20

No other cluster approached that level.

The group also had:

z_{\mathrm{SO/PA}} = -0.32

and:

z_{\mathrm{ISO}} = -0.69

These players produced value through speed, contact, and mobility rather than power.

Central examples include:

• Delino DeShields, 2000
• Stan Javier, 1995
• Marquis Grissom, 1994

Their average OBP was almost exactly equal to the seasonal mean. The cluster was not defined by superior hitting in the narrow sense. A distinctive combination of speed, contact, and limited power defined it.

Did the Archetypes Change Over Time?

This was the study’s central question.

The raw offensive environment changed dramatically. Figure 2 shows a major increase in strikeout rates, higher power levels, and shifting patterns of stolen bases.

The relative distribution of player archetypes was surprisingly stable.

Figure 6. Five-year moving percentages of qualified player-seasons assigned to each cluster.

Patient-contact hitters generally represented between approximately 20 and 28 percent of qualified seasons.

Low-impact contact hitters reached their highest levels during the 1960s and 1970s, then gradually declined. Their five-year moving share peaked near 26 percent in the mid-1970s and stood near 19 percent by 2025.

Power-and-strikeout hitters became somewhat more prevalent, reaching approximately 18 percent around 2012. Yet they never displaced the other archetypes.

Speed-and-contact hitters remained surprisingly consistent. Their moving share generally remained between approximately 8 and 12 percent, even though league-wide stolen-base environments changed considerably.

The elite power-and-patience group remained rare throughout the entire study. It accounted for approximately 8 to 11 percent of qualified seasons in most periods.

The 2020s did not produce a completely new distribution of player types. Compared with the partial 1950s sample, the modern distribution contains somewhat fewer low-impact contact hitters and a modestly larger share of power-and-strikeout hitters.

The basic architecture, however, remains recognizable.

This suggests that baseball evolution has operated on at least two levels.

At the first level, the statistical baseline changes. Strikeouts become more common. Power becomes more valuable. Stolen-base strategies change.

At the second level, players continue to occupy recurring roles relative to that baseline. Every era still contains:

• Patient hitters
• Free swingers
• Power hitters
• Speed specialists
• Low-impact contact hitters
• Rare players who combine several elite skills

The numbers change. The ecological niches persist.

Offensive Archetypes and Defensive Positions

The clusters were also closely connected to defensive position.

Figure 7. Primary defensive positions represented within each offensive archetype.

The elite power-and-patience cluster was concentrated at traditional offensive positions:

• 28.2 percent first basemen
• 17.7 percent right fielders
• 14.1 percent left fielders
• 12.5 percent third basemen

Shortstops represented only 2.1 percent of the cluster.

The power-and-strikeout group followed a similar pattern. First base, right field, and left field accounted for a large portion of those seasons.

The speed-and-contact cluster looked completely different:

• 32.1 percent center fielders
• 20.0 percent shortstops
• 19.7 percent second basemen
• 15.0 percent left fielders

Catchers and first basemen were almost absent.

The low-impact contact group was strongly concentrated in the middle infield:

• 27.9 percent shortstops
• 23.9 percent second basemen

This pattern reflects the interaction between offense and defensive value. A shortstop could remain in a lineup with limited power because his defensive position carried different offensive expectations. A first baseman generally needed much greater offensive production.

The archetypes are therefore not purely hitting categories. They also reflect the way teams distribute offensive and defensive responsibilities across the field.

What the PCA Map Really Shows

The PCA figure is not simply a picture of six boxes.

Instead, it shows a continuous offensive landscape.

The elite power-and-patience hitters occupy the high end of the first principal component because they combine OBP, power, and walks.

The power-and-strikeout hitters move upward on the second component because of their combination of ISO and strikeout rate.

The speed-and-contact hitters move in the opposite direction because their offensive identities are dominated by baserunning and lower power.

The patient-contact and low-impact contact groups overlap along the contact dimension, but separate sharply through OBP and walk rate.

This provides a useful reminder about classification. A player’s archetype is not his complete identity. It is a summary of how his season relates to thousands of other seasons in baseball history.

Some players sit close to a cluster center. Others occupy transitional areas between types.

Limitations

This study has several important limitations.

First, the model uses traditional Lahman statistics. It does not include park-adjusted measures such as wRC+, nor does it include Statcast measures such as exit velocity, barrel rate, launch angle, or sprint speed.

Second, the model treats each player-season as an independent observation. A player who qualified in 15 seasons appears 15 times. This is appropriate for studying the distribution of seasonal styles, but it gives durable players more influence than short-career players.

Third, the cluster labels are interpretations. The algorithm identifies groups of seasons that are statistically similar. It does not name those groups.

Fourth, the sample includes only qualified hitters. Part-time players, platoon specialists, defensive replacements, and many late-career seasons are excluded.

Fifth, seasonal standardization intentionally removes changes in the league-wide baseline. This is the correct approach for identifying relative archetypes, but it means the clusters should not be interpreted as absolute comparisons of offensive production.

A power-and-strikeout hitter from 1960 did not necessarily strike out as frequently as a member of the same cluster in 2025. He struck out frequently relative to the hitters around him.

Conclusion

I began this study expecting to find a succession of offensive types.

I expected contact hitters to dominate the early years, speed players to expand during the 1970s and 1980s, and power-and-strikeout hitters to overwhelm the modern period.

Some of those movements are visible, but the broader result is more interesting.

Baseball’s offensive environment changed enormously. Its fundamental player archetypes changed much less.

The low-strikeout hitter did not disappear. His raw strikeout rate simply rose with the league.

The power-and-strikeout hitter did not suddenly appear in the twenty-first century. Earlier versions existed within lower-strikeout environments.

The speed-and-contact player did not vanish during the power era. His share remained more stable than the raw stolen-base totals might suggest.

Perhaps the evolution of the baseball player is not a story of one species replacing another.

Perhaps it is a story of persistent roles adapting to a changing environment.

The game changes its equilibrium. The players reorganize themselves around it. Yet the same broad offensive strategies continue to reappear, season after season and generation after generation.

That continuity may be one of the most striking features of baseball history.

 

 

Introduction

Home runs are not evenly distributed across the diamond.

That is obvious at one level. First basemen and corner outfielders have historically been expected to hit for power. Middle infielders and catchers have often been evaluated based on a broader mix of defense, contact, arm strength, range, game-calling, and positional scarcity. Designated hitters exist almost entirely because of the bat.

But the obvious pattern is still worth measuring.

This chapter asks a simple question:

How have home runs historically been distributed by primary position?

To answer that, I built a player-season dataset from the Lahman-style batting and appearance files already used in this project. Each player-season was assigned to the position where the player appeared most often. Then I compared home run totals across primary positions.

The main version of the study uses regular player-seasons:

PA \geq 300

That cutoff matters because all player seasons include bench players, call-ups, pitchers, defensive replacements, and partial seasons. Those observations are part of baseball history, but they compress the distribution heavily toward zero. The 300-plate-appearance version gives a clearer view of regular players.

The result is consistent with baseball intuition, but the details are interesting.

Among regular player seasons, the highest median home run totals come from:

DH

1B

RF

LF

3B

The lower median positions are:

SS

2B

C

CF

Third base sits where we might expect it to sit: not quite a pure slugger position like first base, right field, left field, or designated hitter, but clearly more power-oriented than second base or shortstop.

That makes third base a bridge position. It carries defensive responsibility, but it has also historically demanded more power than the middle infield.

Data and Method

The unit of analysis is the player-season.

For each player-season, I summed the player’s batting record across all stints. Home runs came from the batting file:

HR_{i,y} = \sum_{t} HR_{i,y,t}

Where:

i = \text{player} y = \text{season} t = \text{team or stint}

Plate appearances were estimated from the available batting columns as:

PA = AB + BB + HBP + SF + SH

This is the same practical plate-appearance construction used elsewhere in the project.

Each player-season was then assigned a primary position using the appearances file. The primary position was the position at which the player appeared in the most games:

\mathrm{PrimaryPosition}_{i,y} = \operatorname*{arg\,max}_{p} \left( G_{i,y,p} \right)

Where:

G_{i,y,p} = \text{games played by player } i \text{ at position } p \text{ in season } y

The positions included were:

C

1B

2B

3B

SS

LF

CF

RF

DH

P

For the main batting-position figures, I excluded pitchers because pitcher seasons have a very different distribution and can compress the plot.

The main regular-player sample uses:

PA \geq 300

This produced the following number of regular player-seasons by batting position:

C: 2,650

1B: 3,273

2B: 3,273

3B: 3,213

SS: 3,149

LF: 3,262

CF: 3,331

RF: 3,259

DH: 646

The designated hitter sample is smaller because the position did not exist across the full historical record.

Why Use Box Plots?

Home run totals are skewed.

Most players do not hit 40 home runs. Many regulars hit fewer than 10. A smaller number of players produce the spectacular seasons that shape memory and record books.

A box plot is useful because it shows several parts of the distribution at once:

median

middle 50 percent

upper and lower spread

outlier seasons

The median is the central line in the box.

The box itself shows the interquartile range:

IQR = Q_3 - Q_1

Where:

Q_1 = \text{25th percentile} Q_3 = \text{75th percentile}

The outliers show the extreme home run seasons. Those outliers matter because home run history is partly a history of extremes.

The box plot, therefore, helps separate typical power from exceptional power.

Figure 1: Regular Player-Seasons by Primary Position

Figure 1. Home runs by primary position, regular player-seasons.

The first figure is the main result.

It includes batting positions only and uses the regular-player cutoff:

PA \geq 300

The median home run totals are:

C: 7

1B: 12

2B: 5

3B: 9

SS: 4

LF: 10

CF: 7

RF: 11

DH: 18

The pattern is clear.

Designated hitters have the highest median:

\widetilde{HR}_{DH} = 18

First basemen follow:

\widetilde{HR}_{1B} = 12

Right fielders and left fielders are next:

\widetilde{HR}_{RF} = 11 \widetilde{HR}_{LF} = 10

Third basemen sit just behind the corner outfielders:

\widetilde{HR}_{3B} = 9

The middle infield positions are lower:

\widetilde{HR}_{2B} = 5 \widetilde{HR}_{SS} = 4

This is the historical power spectrum in compact form.

The right side of the defensive spectrum contains more home runs. The middle of the diamond contains fewer.

The Third-Base Position in Context

The third-base result is especially relevant to the larger project.

Third base has a median of 9 home runs among regular player-seasons:

\widetilde{HR}_{3B} = 9

Its mean is:

\overline{HR}_{3B} = 11.24

Its 90th percentile is:

P_{90,3B} = 26

Its 95th percentile is:

P_{95,3B} = 32

That means a 26-home-run season by a regular third baseman sits around the 90th percentile historically, while a 32-home-run season sits around the 95th percentile.

This helps frame the third-base offensive studies. A third baseman does not need to hit like a first baseman or designated hitter to be power-relevant at the position. But third base has historically demanded more power than second base or shortstop.

In that sense, third base is not a pure defensive position and not a pure slugger position. It sits between worlds.

Summary Table: Regular Player-Seasons

The main regular-player summary is:

Position   Mean HR   Median HR   90th Percentile   Max

C                    9.20                       7                         20                   60

1B                14.68                       12                       31                   70

2B                 6.89                        5                         16                  45

3B                 11.24                       9                         26                  54

SS                  6.66                       4                          17                  57

LF                 12.47                       10                        27                  73

CF                 10.07                      7                           24                  62

RF                  13.19                      11                          28                 66

DH                19.48                      18                         33                 56

The mean is higher than the median at every position. That reflects the right-skewed nature of home run totals:

\overline{HR}_p > \widetilde{HR}_p

Where:

\overline{HR}_p = \text{mean home runs at position } p \widetilde{HR}_p = \text{median home runs at position } p

This skew is exactly why box plots are useful. The median shows the typical regular. The outliers show the historical power peaks.

Figure 2: All Player-Seasons

Figure 2. Home runs by primary position, all player-seasons.

The second figure includes all player-seasons, regardless of plate appearances.

This plot answers a different question.

Instead of asking what regular players do, it asks what the full population of player-seasons looks like.

The result is much more compressed toward zero.

That is expected.

All player-seasons include players who appeared briefly, bench players, injury-shortened seasons, late-season call-ups, defensive specialists, and players with very few batting opportunities. Many of these seasons have few or no home runs.

This is why the regular-player cutoff is important.

The all-player plot is useful because it shows the structure of the full record. But the regular-player plot is better for comparing positional expectations.

The difference between Figure 1 and Figure 2 is a methodological lesson:

Home run distributions depend heavily on playing-time filters.

That is not a flaw. It is part of the phenomenon.

Figure 3: Including Pitchers

Figure 3. Home runs by primary position, regular player-seasons including pitchers.

The third figure adds pitchers to the regular-player sample.

There are only 49 pitcher seasons with at least 300 estimated plate appearances in this dataset. That is a tiny sample compared with the defensive positions.

Pitchers have a very low home run distribution:

P_n = 49

The pitcher box is compressed near the bottom of the plot.

This is why pitchers are excluded from the main batting-position comparison. Pitchers are not just another defensive position in this context. Historically, they had a very different offensive role.

Including them is still useful as a reminder of how specialized baseball’s offensive expectations have become.

The pitcher comparison also shows why the designated hitter changed the structure of the game. The DH role separated the hitting function from the pitching function.

Figure 4: Before 1994

Figure 4. Home runs by primary position, regular player-seasons before 1994.

The fourth figure examines regular player seasons before 1994.

This view helps separate the long historical baseline from the more recent high-power period.

Before 1994, the positional pattern is still visible:

1B, LF, RF, and DH are higher-power positions.

2B and SS are lower-power positions.

3B sits in the middle-to-upper range.

However, the distributions are generally lower than in the modern period.

That is especially visible in the medians and upper tails. The tops of the boxes and whiskers are lower for most positions.

This matters because the modern home run environment can distort our intuition. If we only look at recent baseball, we may overestimate what a typical power season historically looked like.

The pre-1994 plot keeps the longer history visible.

Figure 5: 1994–2025

Figure 5. Home runs by primary position, regular player-seasons, 1994–2025.

The fifth figure focuses on the modern high-power period from 1994 through 2025.

Here the distributions shift upward.

The median home run totals are visibly higher at most positions than in the pre-1994 plot.

This is particularly clear at positions that historically carried lower power expectations. Second base, shortstop, catcher, and center field all show more power in the modern period than in the older baseline.

The modern plot also shows how the distinction between positions has narrowed in some ways. Shortstops and second basemen now have more home run upside than they did historically, though they still remain below first base and designated hitter in the central distribution.

Third base remains a power-relevant position. In the 1994–2025 period, the third-base distribution shifts upward, reflecting the broader increase in home run production across baseball.

Maximum Home Run Seasons by Primary Position

The maximum regular-player home run seasons by primary position are:

C: Cal Raleigh, 2025, 60 HR

1B: Mark McGwire, 1998, 70 HR

2B: Marcus Semien, 2021, 45 HR

3B: Alex Rodriguez, 2007, 54 HR

SS: Alex Rodriguez, 2002, 57 HR

LF: Barry Bonds, 2001, 73 HR

CF: Aaron Judge, 2022, 62 HR

RF: Sammy Sosa, 1998, 66 HR

DH: Kyle Schwarber, 2025, 56 HR

The maximum values show how different the upper tail can be from the median.

For third base:

\max(HR_{3B}) = 54

The median regular third-base season is:

\widetilde{HR}_{3B} = 9

So the highest third-base season is six times the median:

\frac{ \max(\mathrm{HR}_{\mathrm{3B}}) }{ \widetilde{\mathrm{HR}}_{\mathrm{3B}} } = \frac{54}{9} = 6

This is the shape of home run history. The typical distribution matters, but the record book is made in the tails.

Positional Power Spectrum

The regular-player medians produce a rough power spectrum:

DH > 1B > RF > LF > 3B > C/CF > 2B > SS

This is not a law. It is an empirical summary of the historical data.

It shows that home run expectations follow the defensive spectrum, but imperfectly.

First base and corner outfield positions carry major power expectations. Middle infield positions carry lower typical power expectations. Third base sits between them.

That makes sense. Third base is a reaction position, an arm-strength position, and historically a position where teams have often accepted more offensive responsibility than at shortstop or second base.

Why Third Base Is Interesting

This chapter began as a positional home run study, but it also helps explain why third base is such a useful position for the larger project.

Third base is not first base. It is not a position where the bat alone defines most of the historical expectation.

But it is also not shortstop or second base. The position has carried real offensive expectations for a long time.

This makes third base analytically rich.

A third baseman can be great through power.
A third baseman can be great through defense.
A third baseman can be great through a two-way profile.
A third baseman can be average in one dimension and exceptional in another.

That is why the earlier z-score, WAR, and wRC+ studies were useful. Third base demands a multidimensional approach.

The home run box plots confirm the same idea from a different angle.

Third base lives in the middle of the power spectrum.

Limitations

This study uses primary position by games played. That is practical and transparent, but it simplifies players who split time across multiple positions.

A player-season is assigned to only one position, even if the player spent substantial time elsewhere.

The plate appearance formula is estimated from available batting columns:

PA = AB + BB + HBP + SF + SH

That is a reasonable construction, but it may differ slightly from official plate appearance totals in some historical contexts.

The 300-PA cutoff is also a choice. A different cutoff, such as 250 or 400 PA, would change the sample slightly. The broad pattern should remain, but the exact medians and percentiles would move.

Finally, the designated hitter is not available across the full historical period. It should be interpreted separately from long-running defensive positions.

Conclusion

Home runs have never been distributed evenly across the diamond.

The regular-player box plots show a clear positional structure. Designated hitters, first basemen, right fielders, left fielders, and third basemen occupy the higher-power part of the spectrum. Second basemen and shortstops occupy the lower-power part. Catchers and center fielders sit in between, with major outlier seasons but lower typical medians.

For regular player-seasons:

DH: 18 median HR

1B: 12 median HR

RF: 11 median HR

LF: 10 median HR

3B: 9 median HR

C: 7 median HR

CF: 7 median HR

2B: 5 median HR

SS: 4 median HR

Third base sits in a revealing place.

It is not the highest-power position. It is not the lowest. It is a position where power matters, but not alone.

That is why the position works so well for this larger study.

Third base is a bridge between offensive and defensive expectations.

The home run distributions show that clearly.

Postscript: Where the Home Runs Came From

The box plots in this chapter show the distribution of home run seasons by position. They answer a question about typical and exceptional seasons:

What does a home run season usually look like at each position?

But there is another question worth asking:

Where did the historical home run volume come from?

To answer that, I summed all home runs by primary position.

\mathrm{TotalHR}_{p} = \sum_{i,y} HR_{i,y,p}

Where:

p = \text{primary position}

The first postscript figure uses all player-seasons from 1871 through 2025.

Postscript Figure 1. Total home runs by primary position, all player-seasons, 1871–2025.

The largest total comes from first base:

\mathrm{TotalHR}_{1B} = 56{,}082

Right field and left field follow closely:

\mathrm{TotalHR}_{RF} = 50{,}858 \mathrm{TotalHR}_{LF} = 49{,}898

Third base ranks fourth:

\mathrm{TotalHR}_{3B} = 42{,}140

This reinforces the larger pattern. The historical home run supply has come disproportionately from the corners: first base, corner outfield, and third base.

The second postscript figure uses only regular player-seasons:

PA \geq 300

Postscript Figure 2. Total home runs by primary position, regular player-seasons, PA ≥ 300, 1871–2025.

The regular-player version tells the same basic story. First base remains first:

\mathrm{TotalHR}_{1B} = 48{,}064

Right field remains second:

\mathrm{TotalHR}_{RF} = 42{,}977

Left field remains third:

\mathrm{TotalHR}_{LF} = 40{,}663

Third base remains fourth:

\mathrm{TotalHR}_{3B} = 36{,}127

The designated hitter is the important caution. DH has the highest median home run total among regular player-seasons, but it does not rank near the top in total historical home runs because it is not present across the full 1871–2025 record.

Keep that distinction in mind.

The box plots show rate and distribution.
The bar charts show accumulated historical volume.

They are related, but they are not the same thing.

A position can have a high typical power profile but a smaller historical total if it has fewer seasons in the record. That is exactly what happens with designated hitters.

Third base, by contrast, has both a long historical presence and a meaningful power profile. It does not lead the home run record, but it sits clearly among the power-producing positions.

That is the key postscript result:

First base produced the most historical home run volume.

Corner outfield followed.

Third base remained a major source of home runs.

Designated hitter had the strongest power profile but a shorter historical runway.

 

 

Introduction

At this point in the third-base study, we have three different ways of measuring greatness.

The first is our own z-score framework. It asks how far a player separated from other third basemen in the same season and across a career.

The second is WAR. It asks how much total value a player produced, including offense, defense, baserunning, position, replacement level, and playing time.

The third is wRC+. It asks how strong a hitter was after adjusting for league and park context, with 100 set as league average.

Each measure is useful.

Each measure answers a different question.

That is why this comparison matters.

A player can dominate by z-score because he separates from his third-base peers. A player can dominate by WAR because he accumulates value across many seasons. A player can dominate by wRC+ because his offensive rate quality is extraordinary.

The goal of this chapter is not to declare that one metric is correct and the others are wrong.

The goal is to compare the stories they tell.

For Study 1, I focused on regular third basemen with at least five qualified third-base seasons and matched values for career WAR, career wRC+, and our career z-score measures. That produced a working sample of 239 third basemen.

The central question is:

Which third basemen remain elite when judged by z-scores, WAR, and wRC+ together?

The answer begins with Mike Schmidt.

Across the three-metric composite, Schmidt is the clear anchor of the study. He ranks first in combined z-score, first in WAR, and second in wRC+. He is the player who survives every test.

But the rest of the list is more interesting than a simple ranking.

Eddie Mathews, Chipper Jones, Wade Boggs, George Brett, Home Run Baker, Alex Rodriguez, Ron Santo, Scott Rolen, and Jose Ramirez also emerge as strong cross-metric performers.

At the same time, several players reveal the tension between the metrics.

Brooks Robinson ranks extremely high by combined z-score and WAR, but much lower by wRC+. Dick Allen ranks first by wRC+, but much lower by combined z-score and WAR. Adrian Beltre ranks third by WAR but much lower by wRC+.

Those differences are not problems.

They are the point of the chapter.

The Three Metrics

This study compares three broad dimensions: Combined career z-score, career WAR, and career wRC+.

The combined z-score is internal to this project. WAR and wRC+ are external validation measures.

The three metrics are not interchangeable.

They measure different things.

The Combined Z-Score

The combined z-score is based on two career components: Model C offensive career score  and Traditional defensive career score.

Each career score is standardized across the third-base regular sample.

The standardized offensive score is:

z_{\mathrm{Offense},i} = \frac{ \mathrm{Offense}_{i} - \overline{\mathrm{Offense}} }{ s_{\mathrm{Offense}} }

The standardized defensive score is:

z_{\mathrm{Defense},i} = \frac{ \mathrm{Defense}_{i} - \overline{\mathrm{Defense}} }{ s_{\mathrm{Defense}} }

The combined z-score is:

\mathrm{Combined\ Z}_{i} = z_{\mathrm{Offense},i} + z_{\mathrm{Defense},i}

This score rewards players who separate from other third basemen in both offensive and traditional defensive dimensions.

It is not the same thing as WAR.

It does not directly assign run values. It does not use replacement level. It does not use park factors in the way WAR or wRC+ does. It is a peer-separation measure.

That is its strength.

It asks:

How far did this third baseman stand from the position?

WAR

WAR is a broader value metric.

In this chapter, WAR is used as a career value measure for the player’s qualified third-base seasons in our merged dataset. WAR includes offense, defense, baserunning, positional value, replacement value, and playing time.

For the purpose of this chapter, we can think of WAR abstractly as:

\begin{aligned} \mathrm{WAR}_i &= \mathrm{Offense}_i + \mathrm{Defense}_i + \mathrm{Baserunning}_i \\ &\quad+ \mathrm{Position}_i + \mathrm{Replacement}_i \end{aligned}

That is not intended as a full WAR formula. It is a conceptual summary.

WAR asks a different question from the z-score model.

It asks:

How much total value did this player produce?

That is why Adrian Beltre, Wade Boggs, Brooks Robinson, Scott Rolen, Graig Nettles, and Buddy Bell can look stronger by WAR than they do by wRC+ alone.

WAR values more than hitting.

wRC+

wRC+ is an offensive rate measure.

It is scaled so that 100 is league average:

wRC^+ = 100

A hitter with a 120 wRC+ is roughly 20 percent better than league average offensively:

wRC^+ = 120

A hitter with an 80 wRC+ is roughly 20 percent below league average offensively:

wRC^+ = 80

For this study, wRC+ answers a narrower question:

How good was the hitter?

It does not measure third-base defense. It does not measure total career value. It does not reward playing third base well. It isolates offensive rate quality.

That is why Dick Allen can rank first in wRC+ without ranking first in the other systems.

Why a Composite Score Is Useful

Because the three measures use different scales, we cannot simply add raw z-score, WAR, and wRC+.

Instead, I converted each metric into a percentile rank.

For each player:

P_{\mathrm{Combined\ Z},i} = \mathrm{PercentileRank}(\mathrm{Combined\ Z}_i) P_{\mathrm{WAR},i} = \mathrm{PercentileRank}(\mathrm{WAR}_i) P_{wRC^+,i} = \mathrm{PercentileRank}(wRC^+_i)

Then I calculated a three-metric composite percentile:

\mathrm{Composite}_{i} = \frac{ P_{\mathrm{Combined\ Z},i} + P_{\mathrm{WAR},i} + P_{wRC^+,i} }{3}

This composite is not meant to replace the individual metrics. It is a summary tool.

It rewards broad agreement.

A player who ranks high in all three metrics will rise. A player who is exceptional in one metric but weaker in the others will still be visible, but not necessarily at the top of the composite.

That is why this study is useful.

It separates all-around consensus from metric-specific greatness.

Figure 1: The Top 25 Composite Performers

Figure 1. Top third basemen across combined z-score, WAR, and wRC+.

The top 25 composite chart gives the broadest view of the results.

The top ten are:1. Mike Schmidt 2. Eddie Mathews 3. Chipper Jones 4. Wade Boggs 5. George Brett 6. Home Run Baker 7. Alex Rodriguez 8. Ron Santo 9. Scott Rolen 10. Jose Ramirez.

It is not simply an offensive list. Brooks Robinson does not reach the top ten because wRC+ pulls him down, but the list still includes two-way and value-based players such as Wade Boggs, Ron Santo, and Scott Rolen.

It is not simply a WAR list either. Adrian Beltre ranks third by WAR, but thirteenth by the composite because his wRC+ rank is lower than his WAR and z-score ranks.

It is not simply a wRC+ list. Dick Allen ranks first by wRC+, but he does not land near the top of the composite because his combined z-score and WAR ranks are lower.

The top composite list rewards players who remain strong across the different definitions of greatness.

That is why Schmidt is first.

He is not merely great by one method. He is great by all three.

The Top Players by Each Metric

The top players change depending on the question.

By combined z-score, the top five are: 1. Mike Schmidt 2. Brooks Robinson 3. Nolan Arenado 4. Scott Rolen 5. Wade Boggs.

This list rewards two-dimensional separation. Robinson and Arenado rise because traditional defense is included.

By WAR, the top five are: 1. Mike Schmidt 2. Eddie Mathews 3. Adrian Beltre 4. Wade Boggs 5. Brooks Robinson.

This list rewards total value and career accumulation.

By wRC+, the top five are: 1. Dick Allen 2. Mike Schmidt 3. Eddie Mathews 4. Harmon Killebrew 5. John McGraw.

This list rewards offensive rate quality.

These are three different lists because they are answering three different questions.

The question is not which list is correct.

The question is what each list reveals.

Figure 2: Rank Movement Across the Three Systems

Figure 2. How the top composite third basemen rank by combined z-score, WAR, and wRC+.

The rank-comparison figure shows how players move across the three measures.

Mike Schmidt barely moves. That is the signature of a consensus number one. His profile is not dependent on one definition of value.

Eddie Mathews is similarly strong. He ranks high in WAR and wRC+, and still remains strong in the combined z-score system.

Chipper Jones is also stable. His defensive score is not strong, but his offensive value is so high that he remains near the top.

The movement becomes more interesting with players like Adrian Beltre, Nolan Arenado, and Scott Rolen.

Beltre ranks extremely high by WAR but much lower by wRC+. That makes sense. His case is not purely about offensive rate. It is about durability, defense, and total value.

Arenado ranks very high by combined z-score but much lower by wRC+. Again, that makes sense. His profile is two-dimensional and defense-forward.

Rolen is a balanced case. He ranks very high by combined z-score and WAR but lower by wRC+. That reflects his two-way value.

This figure shows that the metrics are not redundant.

They overlap, but they do not tell the same story.

Figure 3: Combined Z-Score Versus WAR

Figure 3. Combined career z-score versus career WAR among third-base regulars.

The combined z-score and WAR relationship is strong.

The fitted line is:

\mathrm{WAR} = 15.98 + 6.69(\mathrm{Combined\ Z})

The model fit is:

R^2 = 0.782

That means the combined z-score explains a large share of the variation in career WAR among third-base regulars.

This is important.

It tells us that our z-score framework is not just an internal ranking system. It aligns strongly with a major external value metric.

But the scatterplot also shows meaningful differences.

Mike Schmidt sits at the upper-right extreme. His combined z-score and WAR both identify him as historically exceptional.

Brooks Robinson sits high in combined z-score and WAR, but his shape is different. His combined z-score is powered by traditional defense rather than offensive dominance.

Adrian Beltre sits higher in WAR than his combined z-score alone would predict. That suggests his total value, longevity, and broader WAR components are stronger than the simplified z-score model fully captures.

Nolan Arenado sits high in combined z-score but lower in WAR relative to the line. That may reflect career length, active-career status, or differences between traditional defensive separation and WAR’s defensive valuation.

The relationship is strong, but the residuals still matter.

They show where the systems disagree.

Figure 4: Offensive Z-Score Rate Versus wRC+

Figure 4. Average offensive z-score per qualified third-base season versus career wRC+.

The relationship between average offensive z-score and wRC+ is also strong.

The fitted line is:

wRC^+ = 100.89 + 5.41(\mathrm{Average\ Offensive\ Z})

The fit is:

R^2 = 0.740

This confirms the earlier wRC+ validation result.

Average offensive z-score is a strong predictor of wRC+ because both are measuring offensive quality, though in different ways.

The equation says that each additional point of average offensive z-score corresponds to about 5.41 additional points of career wRC+:

\beta_1 = 5.41

This is why the offensive names rise in this figure.

Dick Allen, Mike Schmidt, Eddie Mathews, Chipper Jones, Alex Rodriguez, George Brett, Home Run Baker, Wade Boggs, Al Rosen, and David Wright all appear as strong offensive profiles.

Brooks Robinson, by contrast, is much closer to the middle of the wRC+ distribution. That is not a criticism. It simply reflects that Robinson’s greatness is not primarily a wRC+ case.

That is exactly why this comparison matters.

Figure 5: Cross-Metric Rank Disagreements

Figure 5. Largest cross-metric rank disagreements among notable third basemen.

The disagreement chart is one of the most useful figures in the study.

It identifies players whose rankings differ sharply across combined z-score, WAR, and wRC+.

The rank spread is:

\begin{aligned} \mathrm{RankSpread}_{i} &= \max\left( r_{\mathrm{CombinedZ},i}, r_{\mathrm{WAR},i}, r_{\mathrm{wRC}^{+},i} \right) \\ &\quad- \min\left( r_{\mathrm{CombinedZ},i}, r_{\mathrm{WAR},i}, r_{\mathrm{wRC}^{+},i} \right) \end{aligned}

A large spread means the player looks very different depending on the metric.

Some of the most interesting disagreement cases are: Brooks Robinson, Dick Allen, Adrian Beltre, Nolan Arenado, Willie Kamm, Gary Gaetti, Harmon Killebrew, Edwin Encarnacion, Jim Ray Hart, and Deacon White.

These players are not mistakes in the data.

They are interpretive opportunities.

Brooks Robinson is a defensive and WAR giant, but not a wRC+ giant.

Dick Allen is an offensive-rate giant, but not a top combined z-score or WAR third-base regular in this framework.

Adrian Beltre is a WAR giant, but wRC+ does not fully capture his case.

Willie Kamm is extremely strong by the combined z-score framework because of traditional defense, but he is not similarly high by wRC+.

Edwin Encarnacion is much stronger by wRC+ than by third-base z-score or WAR within the third-base framework, partly because his career offensive identity extends beyond a long regular third-base profile.

The disagreement chart shows why a single number is not enough.

The Schmidt Result

Mike Schmidt is the central result of Study 1.

He ranks: Combined z-score rank: 1 WAR rank: 1 wRC+ rank: 2 Composite rank: 1

This is almost the perfect cross-metric profile.

Schmidt is not merely the best by our internal model. He is also the best by WAR and nearly the best by wRC+.

That matters because it means his result is robust.

He is not a product of one method.

He is the player who remains elite when the question changes.

If the question is peer separation, Schmidt wins.

If the question is total value, Schmidt wins.

If the question is offensive rate quality, Schmidt is still almost at the top.

That is the strongest possible case.

Eddie Mathews, Chipper Jones, and the Offensive Greatness Group

Eddie Mathews ranks second by the composite.

He ranks: Combined z-score rank: 6 WAR rank: 2 wRC+ rank: 3 Composite rank: 2.

That is a very strong cross-metric profile. Mathews does not have Schmidt’s complete separation, but he remains elite everywhere.

Chipper Jones ranks third by the composite: Combined z-score rank: 8 WAR rank: 6 wRC+ rank: 6 Composite rank: 3.

Chipper’s case is offense-forward. His traditional defensive component is not strong, but his offensive quality is so high that he remains elite across the systems.

George Brett and Home Run Baker also belong in this broad offensive greatness group. They are strong by wRC+, strong by WAR, and strong enough by combined z-score to remain near the top.

This group shows that offensive greatness can carry a third-base profile a long way.

Wade Boggs and the On-Base Profile

Wade Boggs ranks fourth by the composite: Combined z-score rank: 5 WAR rank: 4 wRC+ rank: 14 Composite rank: 4.

Boggs is a fascinating case because he is not a home-run power archetype. His greatness is built around contact, on-base skill, batting average, plate discipline, and sustained offensive quality.

The fact that he ranks so highly in the composite confirms that the model is not simply rewarding slugging power.

Boggs was a different kind of offensive star, and the metrics recognize it.

Rolen, Beltre, Arenado, and Two-Way Value

Scott Rolen, Adrian Beltre, and Nolan Arenado show why WAR and combined z-score are necessary companions to wRC+.

Rolen ranks: Combined z-score rank: 4 WAR rank: 10 wRC+ rank: 34 Composite rank: 9.

Beltre ranks: Combined z-score rank: 9 WAR rank: 3 wRC+ rank: 58 Composite rank: 13.

Arenado ranks: Combined z-score rank: 3 WAR rank: 15 wRC+ rank: 57 Composite rank: 15.

These are not weak wRC+ players. But their all-time third-base cases are not primarily wRC+ cases.

They are two-way cases.

Rolen is balanced. Beltre is a total-value and longevity case. Arenado is a defense-forward combined z-score case.

If this study used only wRC+, these players would be underrated.

If it used only WAR, their offensive shape would be less visible.

If it used only z-score, the relationship to broader value would be less clear.

The three-metric comparison gives the fuller picture.

Brooks Robinson and the Limits of wRC+

Brooks Robinson is the clearest example of a player whose greatness is not offensive-rate greatness.

He ranks: Combined z-score rank: 2 WAR rank: 5 wRC+ rank: 114.

That is a huge split.

It makes perfect sense.

Robinson’s historical case is not based on being one of the greatest offensive third basemen. It is based on defense, durability, and total value.

The combined z-score model sees him because traditional defense is included. WAR sees him because total value includes defense. wRC+ does not see him in the same way because wRC+ is an offensive metric.

That is not a flaw in wRC+. It is a reminder that wRC+ answers a narrower question.

Dick Allen and the Limits of Third-Base Accumulation

Dick Allen is the opposite case.

He ranks: wRC+ rank: 1 Combined z-score rank: 87 WAR rank: 51.

Allen’s offensive rate quality is extraordinary. But within this third-base regular framework, he does not accumulate the same kind of third-base-specific z-score or WAR profile as Schmidt, Mathews, Chipper, Boggs, or Brett.

This shows the difference between a great hitter who played third base and a great third baseman across the entire profile. That is an important distinction.

Allen is not diminished by this result. The study simply clarifies what kind of greatness he represents.

He is a wRC+ giant. He is not the top all-around third-base regular by the three-metric composite.

Why This Study Is Interesting

The value of Study 1 is that it prevents the project from becoming metric-dependent.

If Schmidt ranked first only by our z-score model, the conclusion would be interesting but narrower.

But Schmidt also ranks first by WAR and second by wRC+. That makes the conclusion much stronger.

At the same time, the disagreements prevent the chapter from becoming too simple.

Brooks Robinson, Dick Allen, Adrian Beltre, Nolan Arenado, Rolen, Boggs, and others show that greatness has different forms.

The study therefore supports two conclusions at once: 1. Mike Schmidt is the clearest cross-metric third-base anchor. 2. Different metrics reveal different kinds of third-base greatness.

Both points are important.

Limitations

This chapter uses regular third basemen with at least five qualified third-base seasons and matched values across the three systems. That makes the comparison cleaner, but it also means the study is focused on third-base regulars, not every player who ever appeared at third base.

The combined z-score uses this project’s offensive Model C and traditional defensive model. It does not include modern defensive metrics, park factors, or full run-value modeling.

WAR includes many components that the z-score model does not.

wRC+ is a rate statistic and should not be treated as an accumulated career value measure. That is why the study includes average offensive z-score per qualified season when comparing to wRC+.

The composite percentile score is a summary tool. It is not a new definitive metric. It is best used to identify players who remain strong across multiple systems.

Conclusion

Study 1 compares three ways of measuring third-base greatness: Combined z-score, WAR, and wRC+

The main result is clear.

Mike Schmidt is the strongest cross-metric third baseman in the study.

He ranks first by combined z-score, first by WAR, second by wRC+, and first by the three-metric composite.

Eddie Mathews, Chipper Jones, Wade Boggs, George Brett, Home Run Baker, Alex Rodriguez, Ron Santo, Scott Rolen, and Jose Ramirez also emerge as strong cross-metric performers.

But the disagreements are just as important.

Brooks Robinson shows that wRC+ cannot capture defensive greatness.

Dick Allen shows that offensive rate greatness is not the same as all-around third-base accumulation.

Adrian Beltre, Nolan Arenado, and Scott Rolen show the importance of two-way value.

The larger conclusion is this:

Third-base greatness is not one-dimensional.

Z-scores show peer separation.
WAR shows total value.
wRC+ shows offensive quality.

The best third basemen are the ones who remain visible when the lens changes.

By that standard, Mike Schmidt stands at the center of the argument.

 

The earlier chapter asked a ranking question: which MLB organizations built the best pitching staffs from 2001 to 2025?

This post asks a slightly different question.

Which teams pitched alike?

That is not the same thing. Two teams can both be good without being similar. One team might dominate through strikeouts and fielding-independent indicators. Another might prevent runs through contact control, ground balls, park fit, or bullpen management. A third might have a mixed profile, with decent run prevention but weaker underlying indicators. Ranking tells us who was best. Clustering tells us who had the same shape.

That is why a dendrogram is useful. It does not begin with a leaderboard. It begins with resemblance.

The question becomes: if we describe each franchise by its long-term pitching profile, which franchises naturally group together?

The method

Each franchise was described using season-normalized pitching variables from 2001 through 2025. That step is important because pitching environments changed dramatically during this period. A 4.00 ERA in 2001 does not mean exactly the same thing as a 4.00 ERA in 2025.

So each team-season was first compared to its own season.

z_{i,y,m} = \frac{ X_{i,y,m} - \overline{X}_{y,m} }{ s_{y,m} }

Here, (X_{i,y,m}) is team (i)’s value for metric (m) in season (y), (\bar{X}_{y,m}) is the league average for that metric in that season, and (s_{ym}) is the season standard deviation.

For lower-is-better metrics, such as ERA-, FIP-, xFIP-, SIERA, BB%, HR/9, HR/FB, and Hard%, I reversed the sign so that higher values always mean a more favorable pitching profile.

The clustering used these long-term franchise traits:

Category	Variables
Value	WAR
Dominance	K-BB%
Command	BB% prevention
Run prevention	ERA-
Fielding-independent skill	FIP-, xFIP-, SIERA
Home-run control	HR/9, HR/FB
Contact profile	GB%, Hard% prevention
Starter usage	Quality-start rate

After calculating each franchise’s average profile, I standardized the franchise-level variables and used Ward hierarchical clustering. The distance between teams is based on how far apart their standardized pitching profiles are.

d(i,j) = \sqrt{ \sum_{m=1}^{p} \left( z_{i,m} - z_{j,m} \right)^2 }

The dendrogram then links the most similar teams first and gradually joins them into larger groups.

Figure 1. MLB team pitching identity dendrogram, 2001-2025

The first thing to notice is that the dendrogram is not only a quality ranking. It does separate many of the best pitching organizations, but it also captures style.

The Dodgers, Yankees, Astros, Guardians, Phillies, Cubs, and Red Sox form a major cluster. That makes sense. These organizations score well across the strongest modern pitching indicators: WAR, K-BB%, FIP-, xFIP-, and SIERA. This is the elite skill-and-value cluster.

But the Braves, Giants, and Cardinals form a different group. They are not grouped with the Dodgers and Yankees, even though they include strong pitching organizations. Their similarities lie more in run prevention, home-run suppression, ground-ball tendency, and contact management. In other words, their profile is not merely “good pitching.” It is a particular kind of good pitching.

The Rays, Brewers, Padres, Blue Jays, and Diamondbacks form another interesting group. This cluster is more modern and peripheral-driven. These teams tend to show some strength in strikeout-minus-walk skill and xFIP/SIERA-style indicators, but they are not as dominant in long-term value or run prevention as the elite group.

At the other end, the Rockies, Royals, Orioles, Reds, and Rangers group together as long-term struggling pitching profiles. That does not mean each franchise was bad every year. It means that across the full 2001-2025 window, their average profile shares several weaknesses: lower WAR, weaker K-BB%, weaker fielding-independent indicators, and poorer home-run prevention.

The six main clusters

The dendrogram produced six useful interpretive groups:

Cluster	Teams	Interpretation
Elite skill and value staffs	BOS, CHC, CLE, HOU, LAD, NYY, PHI	Strongest overall skill profile, especially WAR, K-BB%, FIP-, xFIP-, and SIERA
Contact-control run preventers	ATL, SFG, STL	Strong run prevention, home-run control, ground-ball tendency, and starter length
Modern peripheral builders	ARI, MIL, SDP, TBR, TOR	Better in K-BB%, xFIP, and SIERA than in long-term WAR or ERA dominance
Mixed middle profiles	ATH, CHW, DET, LAA, MIN, NYM, SEA, WSN	No single shared identity as strong as the other groups, generally middle-range profiles
Low-dominance HR suppressors	MIA, PIT	Weak dominance indicators but relatively better home-run suppression
Long-term struggling profiles	BAL, CIN, COL, KCR, TEX	Broadly weak long-term pitching profile across value, dominance, and fielding-independent metrics

Why the clusters formed

Figure 2 explains the dendrogram. It shows the average profile of each cluster.

The elite skill-and-value group is strong almost everywhere that modern pitching analysis would expect. The group is especially strong in K-BB%, FIP-, xFIP-, SIERA, and WAR. This is the clearest “modern excellence” cluster.

The contact-control group is different. Atlanta, San Francisco, and St. Louis are not grouped primarily because of overwhelming strikeout dominance. They cluster because of run prevention, home-run prevention, ground-ball tendency, and starter length. This is a more traditional-looking run-prevention cluster.

The modern peripheral group is subtle. Arizona, Milwaukee, San Diego, Tampa Bay, and Toronto do not dominate the long-term value category. Still, they show enough similarity in strikeout-minus-walk skill and advanced indicators to cluster together. This feels like a group of organizations that, at different times, leaned into modern pitching design without producing the same full-period dominance as the Dodgers or Yankees.

The struggling group is also clear. Baltimore, Cincinnati, Colorado, Kansas City, and Texas sit below average in most of the key categories. Colorado’s presence is not surprising because pitching in Denver creates a unique environmental problem. But the dendrogram is not only about park effects. The broader cluster reflects weaker long-term skill indicators too.

Why the Dodgers separate

The Dodgers are inside the elite skill-and-value cluster, but they remain visually distinctive. They join the group later than several other teams inside that cluster. That is important.

It suggests that the Dodgers are similar to other elite pitching organizations but are still somewhat their own thing. Their long-term profile is so strong across so many categories that they are not simply interchangeable with the Yankees, Astros, Guardians, Phillies, Cubs, or Red Sox.

That matches the earlier chapter. The Dodgers were the top pitching organization by average normalized score from 2001 to 2025. The dendrogram confirms the same story from a different angle. They are not only highly ranked. They have a recognizable organizational profile.

Houston and Cleveland as development stories

Houston and Cleveland are also revealing.

The Astros cluster with the elite organizations, but their full-period story contains more transformation than the Dodgers’ story. Houston’s early-2010s pitching collapse and late-2010s pitching rise are both part of the same 25-year average. Even with those bad seasons included, Houston still clusters with the strongest pitching organizations. That says something about how powerful the later Houston pitching model became.

Cleveland’s placement also makes sense. The Guardians are not always discussed like the Dodgers or Yankees because they operate with a different market profile, but the data places them in the same broad pitching family. Their identity is built around development, strike-zone control, and repeatable pitching skill.

Tampa Bay, Milwaukee, and the modern middle

The Rays and Brewers are especially interesting because they appear in the modern peripheral group rather than the elite skill-and-value group.

That is not an insult. It may actually be the more interesting result.

Tampa Bay and Milwaukee often represent modern pitching creativity: bullpen flexibility, role adaptation, pitcher development, and tactical staff construction. But over the full 25-year period, the dendrogram places them closer to San Diego, Toronto, and Arizona than to the Dodgers or Yankees.

That suggests a distinction between modernity and dominance. A team can be tactically modern without producing the same long-term value profile as the very top organizations.

What the dendrogram adds

The dendrogram adds something that a leaderboard cannot.

A leaderboard says:

The Dodgers were first.
The Yankees were second.
The Astros, Guardians, Cubs, Red Sox, Phillies, and Braves followed.

The dendrogram says something different:

The Dodgers, Yankees, Astros, Guardians, Cubs, Red Sox, and Phillies belong to a shared skill-and-value family.

The Braves, Giants, and Cardinals form a separate contact-control and run-prevention family.

The Rays, Brewers, Padres, Blue Jays, and Diamondbacks form a modern peripheral family.

The Rockies, Royals, Orioles, Reds, and Rangers share a long-term struggling profile.

That is the value of clustering. It changes the question from “who was better?” to “who was built alike?”

Conclusion

The central lesson is that pitching identity has shape.

Some organizations built strong staffs across nearly every modern indicator. Some built staffs that prevented runs through contact management and home-run control. Some had modern peripheral strengths without the same long-term value profile. Some struggled across nearly every measure.

The dendrogram does not replace the chapter’s rankings. It deepens them.

From 2001 to 2025, the best pitching organizations were not merely collecting arms. They were building systems. The Dodgers built the most consistent system. The Yankees, Astros, Guardians, Phillies, Cubs, and Red Sox clustered near them because they also produced strong long-term skill profiles. The Braves, Giants, and Cardinals showed a different path, one built more around run prevention and contact control.

The larger point is simple: team pitching is not just performance. It is identity.

And the dendrogram shows us how identity takes shape.

 

 

Pitching is one of the most difficult parts of baseball to measure because it sits at the intersection of many things. A pitcher controls the ball, but not everything that happens after contact. A defense turns balls in play into outs, or fails to. A park changes the meaning of a fly ball. A league environment changes the meaning of a 4.00 ERA. A bullpen changes the way we understand a starter. A front office changes the way we understand a pitching staff.

That is why a long-term team pitching study has to be deliberate and careful. If we simply rank every team from 2001 to 2025 by ERA, we are mixing together very different run environments. A 3.70 team ERA in one season does not mean exactly the same thing as a 3.70 team ERA in another. The offensive environment changes. The baseball changes. Strikeout rates change. Bullpen usage changes. Even the definition of a normal starting pitcher changes.

So the goal of this chapter is not merely to ask which team had the lowest ERA. The better question is this: which organizations consistently produced strong pitching staffs relative to their own era?

That is a critical distinction. A team does not pitch in the abstract. It pitches in a particular season, against a particular league, with a particular baseball, inside a particular tactical environment. The 2001 Diamondbacks, the 2011 Phillies, the 2017 Guardians, the 2018 Astros, and the 2024 Braves all belong to the same broad story, but they do not belong to the same pitching world.

The data in this study covers MLB team pitching from 2001 through 2025. The core variables include ERA, ERA-, FIP, FIP-, xFIP, xFIP-, SIERA, WAR, K%, BB%, K-BB%, HR/9, HR/FB, complete games, quality starts, and several contact-profile measures. For 2001, xFIP, SIERA, and detailed contact data are incomplete, so the 2001 season is included in the main study but handled carefully where those variables are missing.

The central finding is straightforward: from 2001 to 2025, team pitching moved decisively toward strikeout-based run prevention. The best organizations were not simply the ones that prevented runs in a given season. They were the ones who repeatedly built staffs with strong strikeout-minus-walk rates, strong fielding-independent indicators, and enough depth to remain competitive across changing offensive environments.

The Dodgers stand out most clearly. Over the full 25-year period, they were the strongest pitching organization by average normalized pitching score. The Yankees, Astros, Guardians, Cubs, Red Sox, Phillies, and Braves also appear near the top. But the Dodgers are the outlier, not because of one spectacular season, but because of repeated organizational excellence.

Method: comparing teams within seasons

The first methodological problem is that pitching statistics are unstable over time. A league-average pitching staff in 2001 did not look like a league-average pitching staff in 2025. Strikeouts increased. Complete games declined. Velocity rose. Bullpen usage expanded. Home-run rates surged and retreated. If we compare raw numbers across all years, we risk confusing historical context with team quality.

To solve this, each team-season was compared only to the other teams from the same season. In other words, the 2011 Phillies were compared to the league as a whole in 2011. The 2018 Astros were compared with the league in 2018. The 2024 Braves were compared to the league as a whole. This allows us to ask which staffs were exceptional relative to the environment in which they actually pitched.

The basic within-season z-score is:

z_y(X_{i,y}) = \frac{ X_{i,y} - \overline{X}_y }{ s_y(X) }

Here, ( X_{i,y}) is a statistic for team (i) in season (y),  (\overline{X}_y) is the league average for that statistic in that season, and (s_y(X)) is the standard deviation across teams in that season.

For statistics where higher is better, such as WAR and K-BB%, the z-score is used directly. For statistics where lower is better, such as ERA-, FIP-, xFIP-, and SIERA, the sign is reversed. This keeps the interpretation consistent. A higher score always means better pitching.

q_{i,y,m} = \begin{cases} z_y(m_{i,y}), & \text{if higher values are better} \\ -z_y(m_{i,y}), & \text{if lower values are better} \end{cases}

The overall pitching score is then the average of the available component scores:

\text{Pitching Score}_{i,y} = \frac{1}{|M_{i,y}|} \sum_{m \in M_{i,y}} q_{i,y,m}

The metric set is:

M = \left\{ \text{WAR}, \text{K-BB\%}, \text{ERA-}, \text{FIP-}, \text{xFIP-}, \text{SIERA} \right\}

This score is not meant to be the only possible definition of pitching quality. It is a deliberately balanced measure. It includes actual run prevention through ERA-, fielding-independent performance through FIP- and xFIP-, skill-based dominance through K-BB%, and overall value through WAR.

K-BB% is especially important because it captures the two plate appearance outcomes most directly controlled by the pitcher: strikeouts and walks.

\text{K-BB\%} = \text{K\%} - \text{BB\%}

FIP also deserves special attention because it attempts to isolate the events most directly connected to the pitcher: home runs, walks, hit batters, and strikeouts.

\text{FIP} = \frac{ 13 \cdot \text{HR} + 3 \cdot (\text{BB} + \text{HBP}) - 2 \cdot \text{K} }{ \text{IP} } + c_{\text{FIP}}

The constant ( c_{\text{FIP}} ) places FIP on an ERA-like scale. Because that run environment changes by season, FIP- and other indexed measures help compare teams more fairly.

The league changes: strikeouts become the center of pitching

The first figure shows the most important league-wide transformation: the rise of strikeouts.

Figure 1. K%, BB%, and K-BB% trends, 2001-2025

In 2001, the league strikeout rate was about 17.3%. By 2025, it was about 22.2%. That is not a small tactical adjustment. It is a structural change in how pitching works. The modern pitching staff is built around missing bats in a way that the early 2000s staff was not.

Walk rate did not change nearly as dramatically. In 2001, the league walk rate was about 8.4%. In 2025, it was about 8.4% again. There were fluctuations in between, but the broad pattern is clear: strikeouts rose much more than walks did.

That means K-BB% increased substantially. In 2001, league K-BB% was about 8.9%. In 2025, it was about 13.8%. That is the heart of the modern pitching revolution. The best staffs are not just striking out more hitters. They are increasing the gap between strikeouts and walks.

This is why K-BB% belongs near the center of the composite score. It is simple, but powerful. It strips pitching down to a basic contest: can the staff create strikeouts without giving back too many free baserunners?

The answer, increasingly, is yes. But not all teams answered equally well.

Run prevention, FIP, and the changing meaning of ERA

The second figure compares ERA, FIP, xFIP, and SIERA over time. Note that some of the data overlaps on the same line.

Figure 2. ERA, FIP, xFIP, and SIERA trends, 2001-2025

One of the striking features of the long-term data is that ERA does not move in one simple direction. The league ERA was about 4.42 in 2001. It fell to about 3.74 in 2014, rose to about 4.51 in 2019, and then settled around 4.16 in 2025.

That pattern matters because it reminds us that pitching quality cannot be evaluated solely by raw ERA. A team can have a lower ERA because it is genuinely better, but it can also have a lower ERA because the entire league is scoring less. Likewise, a higher ERA in a high-offense environment may not be as bad as it looks.

The 2014 season is a useful example. League run prevention was strong. ERA, FIP, xFIP, and SIERA all sat at relatively low levels. A good pitching staff in 2014 had to be judged against that lower-scoring context. The opposite problem appears in 2019, when the home-run environment pushed run scoring upward. A team that survived 2019 with strong FIP-based indicators deserves credit, given the more difficult environment.

That is why indexed statistics such as ERA- and FIP- are valuable. They tell us how a staff performed relative to league average, where lower is better.

\text{ERA-} = 100 \cdot \frac{ \text{Team ERA} }{ \text{League ERA} } \quad \text{adjusted for context}

A team with an ERA- of 90 was roughly 10% better than league average by that measure. A team with an ERA- of 110 was roughly 10% worse. The same logic applies to FIP-, except the foundation is fielding-independent pitching rather than actual runs allowed.

This distinction becomes crucial when comparing staffs across 25 seasons. The 2011 Phillies and the 2018 Astros both appear as historically great staffs, but they do not look great in exactly the same way. The Phillies represent a more traditional elite rotation model. The Astros represent the modern strikeout, command, and run-prevention model.

The home-run environment

The third figure shows the home-run environment.

Figure 3. HR/9 and HR/FB trends, 2001-2025

Home runs are one of the most important pressure points in modern pitching analysis because they connect individual pitcher skill, batted-ball profile, park context, and league environment. In 2001, league HR/9 was about 1.14. It dipped below 1.00 in several seasons, including 2010 and 2014, then spiked dramatically in 2019, reaching about 1.41 HR/9.

The 2019 season stands out immediately. It was not merely a season with more scoring. It was a season in which the relationship between contact and damage changed. HR/FB also rose sharply, reaching about 15.3% in 2019. That created a very different environment for pitchers.

This is one reason xFIP can be useful. FIP uses actual home runs allowed. xFIP estimates performance by normalizing home-run rate relative to fly balls. Neither statistic is perfect. FIP gives pitchers the actual cost of the home runs they allowed. xFIP asks whether that home-run rate was likely to persist.

For a team-level study, the difference between FIP and xFIP can be revealing. A team with a much better FIP than xFIP may have suppressed home runs unusually well, perhaps through park effects, pitcher skill, batted-ball management, or some combination of these. A team with a much worse FIP than xFIP may have been punished by an elevated home-run rate.

The home-run environment also helps explain why season-normalization is necessary. A team pitching in 2019 faced a different kind of run-prevention problem than a team pitching in 2014. The raw numbers alone cannot tell us whether a staff was good. They have to be interpreted against the league context.

The disappearance of the complete game

The fourth figure captures one of the clearest tactical changes in baseball.

Figure 4. Complete games and quality starts, 2001-2025

In 2001, MLB teams combined for 199 complete games. In 2002, that number was 214. By 2025, it had fallen to 29.

This is not a gradual stylistic preference. It is a transformation in pitcher usage. The complete game went from a normal, if still special, part of pitching to a rare event. The starting pitcher’s job changed. The bullpen’s job changed. The manager’s job changed. The entire architecture of run prevention changed.

Quality starts also declined. In 2001, teams combined for 2,342 quality starts. In 2025, that number was 1,676. Unlike complete games, quality starts did not nearly vanish; they simply became less central to how team pitching is organized.

This matters because a traditional pitching staff was often understood through the front of the rotation. The ace mattered. The number two starter mattered. The innings-eater mattered. In the modern game, those categories still matter, but they are less complete descriptions of team pitching quality. A staff can be excellent because it has dominant starters, but it can also be excellent because it has a deep bullpen, matchup flexibility, velocity, strikeout depth, and player-development infrastructure.

This is one reason team-level pitching analysis is valuable. It captures the staff as an organization, not merely as a list of starting pitchers.

The organizational scoreboard

The franchise-level results show which organizations repeatedly built strong pitching staffs across the full period.

Figure 5. Franchise average pitching score, 2001-2025

The top organizations by average normalized pitching score were:

Rank	Franchise	Avg Score	Avg Rank	Top-5 Seasons
1	LAD	1.043	5.80	16
2	NYY	0.785	8.12	10
3	HOU	0.470	11.20	8
4	CLE	0.447	12.16	9
5	CHC	0.363	12.80	6
6	BOS	0.361	12.44	6
7	PHI	0.356	11.88	6
8	ATL	0.344	12.60	7

The Dodgers are the clear leader. Their average rank was 5.80 across 25 seasons, and they finished in the top five 16 times. That is a remarkable level of consistency.

The Yankees also stand out. They were not as dominant as the Dodgers by average score, but they were consistently strong. Their average rank was 8.12, and they had 10 top-five seasons.

Houston’s position is interesting because the Astros’ 25-year period includes both very bad years and elite years. Their full-period average ranks third, but that average hides a sharp organizational transformation. The 2013 Astros appear among the worst pitching seasons in the dataset. The 2018 and 2019 Astros appear among the strongest. That makes Houston one of the most dramatic before-and-after stories in the study.

Cleveland also deserves attention. The Guardians were not merely good in one season. They produced nine top-five seasons across the full period, including the remarkable 2017 staff and the shortened-season 2020 staff. Cleveland’s results point toward a consistent ability to develop or acquire pitching skill, especially strikeout and command skill.

The Phillies are different. Their full-period average is strong, but their story is anchored by the 2011 staff, the top single-season result in the study. The Phillies’ score is not just about consistency. It is about peak excellence.

The best team pitching seasons

The best individual team pitching seasons in the study were:

Figure 6. Best team pitching seasons, 2001-2025

Rank	Season	Team	Score	WAR	ERA-	FIP-	K-BB%
1	2011	PHI	2.472	29.45	78.58	82.88	14.75%
2	2017	CLE	2.464	30.35	72.49	75.44	20.59%
3	2018	HOU	2.442	28.63	75.89	78.23	21.17%
4	2013	DET	2.089	26.26	89.31	82.13	15.79%
5	2024	ATL	2.059	23.62	84.13	86.61	18.47%

The top three are especially revealing because they show three different versions of elite pitching.

The 2011 Phillies represent the great traditional staff. Their rotation was the center of the story. Their ERA- was 78.58, meaning they were far better than league average at preventing runs. Their FIP- was also excellent at 82.88. They were not merely outperforming their peripherals. They were genuinely strong across the major indicators.

The 2017 Guardians look like a bridge between traditional excellence and modern dominance. Their ERA- was 72.49, the best among the top five listed here, and their FIP- was 75.44. Their K-BB% was 20.59%, which is extraordinary. This is a staff that combined run prevention, fielding-independent strength, and strikeout-minus-walk dominance.

The 2018 Astros are the modern model. Their K-BB% was 21.17%, the highest among these top five. Their ERA- and FIP- were both outstanding. They did not merely prevent runs. They controlled the plate appearance.

That phrase may be the key to the whole chapter: the modern elite staff controls the plate appearance. It wins by turning fewer balls into uncertain events. More strikeouts. Fewer walks. Better home-run control. Better matchup deployment. Better depth.

The 2013 Tigers are also fascinating. Their FIP- was much stronger than their ERA-, which suggests a staff whose fielding-independent indicators were better than its actual run prevention. That kind of gap is analytically useful because it may point toward defense, sequencing, bullpen leakage, park effects, or simple variation.

The 2024 Braves round out the top five, showing that the modern model remains alive. Strong WAR, strong ERA-, strong FIP-, and excellent K-BB% place them among the best team pitching seasons of the last 25 years.

The heat map view: organizational memory

The heat map shows how pitching strength is distributed over time.

Figure 7. Team pitching score heat map, 2001-2025

A heat map is useful because it shows continuity. A table gives us leaders. A heat map gives us memory.

The Dodgers’ consistency becomes visible immediately. They do not merely spike and disappear. They remain strong across many different league environments. This suggests that their pitching success is not just the product of one rotation or one era. It is organizational.

The Yankees also show long-term strength, although with a different shape. They remain regularly above average, but the Dodgers’ top-end consistency is stronger.

Houston’s pattern is more dramatic. The Astros transitioned from poor pitching during the rebuilding years to elite pitching in the late 2010s and beyond. This makes them one of the best examples of organizational reinvention in the dataset.

Cleveland’s pattern is also compelling. The Guardians do not always (understatement) have the resources of the largest-market teams, but the pitching results are consistently strong enough to suggest a real developmental identity. Cleveland’s peak seasons are not accidents.

Tampa Bay deserves separate attention as well. The Rays do not rank at the very top over the full period, but their modern pitching identity is clear. They are one of the organizations most associated with bullpen creativity, opener usage, and flexible staff construction. A team-level study captures some of that, although a starter-reliever split would make the story even sharper.

While the heat map shows the full league, a smaller set of franchise trajectories makes the organizational story easier to see. Figure 8 follows several teams that help define the period: the Dodgers, Yankees, Astros, Guardians, Rays, Phillies, and Braves. The Dodgers show sustained excellence. Houston shows dramatic organizational reinvention. Cleveland and Tampa Bay show the value of pitching development and tactical adaptation. Philadelphia shows the difference between peak rotation dominance and long-term consistency.

Figure 8. Selected Franchises, 2001 – 2025

Three eras of team pitching

Breaking the study into periods helps clarify the historical movement.

From 2001 to 2009, the leading organizations were:

Period	Rank	Franchise	Avg Score	Avg Rank	Top-5 Seasons
2001-2009	1	CHC	0.980	6.89	4
2001-2009	2	LAD	0.840	7.33	5
2001-2009	3	ARI	0.805	8.78	5
2001-2009	4	BOS	0.791	8.22	4
2001-2009	5	NYY	0.733	9.00	4

The early period is more rotation-centered. The Cubs, Dodgers, Diamondbacks, Red Sox, and Yankees all had strong stretches. This era still belongs partly to the older model of staff construction. Starting pitching carries more of the symbolic weight. Complete games are declining, but they have not yet collapsed to modern levels.

From 2010 to 2019, the leaders were:

Period	Rank	Franchise	Avg Score	Avg Rank	Top-5 Seasons
2010-2019	1	LAD	1.202	4.40	7
2010-2019	2	NYY	0.925	6.70	4
2010-2019	3	WSN	0.733	9.20	3
2010-2019	4	CLE	0.709	10.20	6
2010-2019	5	TBR	0.687	8.60	2

This is the period when the modern pitching environment becomes much clearer. Strikeouts rise. Velocity rises. Bullpen roles become more specialized. Cleveland, Tampa Bay, Washington, and Los Angeles all become central parts of the story.

From 2020 to 2025, the leaders were:

Period	Rank	Franchise	Avg Score	Avg Rank	Top-5 Seasons
2020-2025	1	LAD	1.082	5.83	4
2020-2025	2	PHI	0.941	5.67	3
2020-2025	3	MIL	0.808	8.00	2
2020-2025	4	TBR	0.808	6.83	2
2020-2025	5	ATL	0.658	11.33	2

The modern period is especially interesting because it includes the shortened 2020 season, the post-2020 workload reset, and the continuing dominance of strikeout-based staff construction. The Dodgers remain first. The Phillies rise. The Brewers and Rays become central examples of modern pitching development and staff management. The Braves also emerge strongly, especially with the 2024 season.

The worst seasons and the cost of weak pitching infrastructure

The worst team pitching seasons are just as revealing as the best ones.

At the bottom of the dataset are seasons such as the 2025 Rockies, 2006 Royals, 2023 Athletics, 2024 Rockies, and 2013 Astros. These seasons combine weak WAR, poor run prevention, poor FIP-based indicators, and low K-BB%.

The 2025 Rockies had a pitching score of -2.757, with a 125.19 ERA-, 119.75 FIP-, and only an 8.47% K-BB%. The 2006 Royals were similarly poor, with a 124.43 ERA-, 118.20 FIP-, and a 4.15% K-BB%. The 2023 Athletics had a 132.94 ERA-, 122.15 FIP-, and 9.49% K-BB%.

These are not merely bad ERAs. They are broad staff failures. When a team is poor in both run prevention and fielding-independent indicators, the problem is deeper than sequencing or defense. It suggests that the staff is not controlling the strike zone, not limiting damaging contact enough, and not producing enough value.

The 2013 Astros are especially important because they later became one of the strongest pitching organizations in the study. That contrast gives us a natural case study in organizational transformation. Bad pitching staffs do not have to remain bad forever. But the transformation requires more than one good pitcher. It requires a system.

What the study suggests

This first pass suggests several conclusions.

First, team pitching from 2001 to 2025 became increasingly strikeout-centered. The rise in K% and K-BB% is the central statistical movement of the period. It changed what good pitching looks like.

Second, raw ERA is not enough for a long-term study. ERA remains important because runs allowed are real. But ERA must be placed next to FIP, xFIP, SIERA, K-BB%, and indexed measures such as ERA- and FIP-. Otherwise, we risk mistaking league environment for team quality.

Third, complete games and traditional starter workload declined dramatically. This changes how we should think about team pitching. A great staff is no longer just a great rotation. It is a complete run-prevention system.

Fourth, the Dodgers are the strongest pitching organization of the 2001-2025 period. Their dominance is not just peak dominance. It is consistency. They averaged a top-six pitching rank across 25 seasons and finished in the top five 16 times.

Fifth, several organizations deserve deeper case studies. The Astros show organizational reinvention. The Guardians show player-development strength. The Rays show tactical creativity. The Phillies show the power of peak rotation excellence. The Braves show modern staff strength. The Yankees show long-term high-level stability.

Conclusion: pitching as organizational identity

The most important lesson from this study is that pitching is no longer best understood as a collection of individual arms. At the team level, pitching has become an organizational identity.

The best teams do not merely find pitchers. They shape pitching environments. They develop velocity. They manage workloads. They build bullpens. They optimize matchups. They control the strike zone. They use data to turn raw stuff into repeatable advantage.

That is why the Dodgers’ long-term record matters. It is not just that they had good pitchers. Many teams have good pitchers for a year or two. The Dodgers repeatedly built strong pitching staffs across different run environments, tactical eras, and roster cycles.

The same broader lesson applies to Houston, Cleveland, Tampa Bay, Milwaukee, Atlanta, Philadelphia, and New York. The details differ, but the underlying pattern is the same. Modern pitching excellence is systemic.

From 2001 to 2025, baseball shifted toward a game in which the best staffs increasingly controlled plate appearances. Strikeouts rose. Walks became more costly. Home runs reshaped risk. Complete games disappeared. Bullpens expanded. The old image of pitching as one starter carrying a game into the ninth inning gave way to something more distributed, more specialized, and more organizational.

The great pitching staffs of this period are therefore not just statistical outliers. They are historical markers. They show how the game changed, and how the smartest organizations changed with it.