/cdn.vox-cdn.com/uploads/chorus_image/image/50158989/921989.0.jpg)
Joe DiMaggio is in the news again. It’s been 75 years since he concluded his historic 56-game hitting streak with the Yankees, thus cementing his place in baseball history. The Yankees tweeted out a "news break" every day this summer, updating followers on where Joltin’ Joe was during the streak. Fans could re-live the feat in real time, as it historically unfolded.
Brian Kenny writes extensively about DiMaggio’s 1941 season in his new book, Ahead of the Curve, laying out the statistical case that in fact, the baseball writers voted for the right player that year for the American League’s Most Valuable Player award. See, Ted Williams of the Red Sox has been the cognoscenti’s darling in modern decades, thanks to superior numbers in those fancy categories like OPS, wRC+, and WAR. We’ve even argued Williams’s case before in these pages. It would seem like Teddy Ballgame’s 1941 season was a classic instance of sportswriter injustice, the BBWAA writers whiffing on a deserving candidate in the face of a compelling narrative.
Kenny, however, advocates for Joltin’ Joe, showing that the Yankee Clipper was a more valuable player than Williams in that pre-Pearl Harbor year, thanks to context-dependent numbers like Win Probability Added and Run Expectancy. Kenny:
Even with Ted Williams having the then best on-base percentage of all time, Joe DiMaggio did more to score more runs for his team. When you add up all of his at bats [sic] and how he changed the probability of scoring runs, DiMaggio was better. [...]
In hitting alone, not counting defense, DiMaggio changed the probability of his team winning more than Williams did. The game-winning ability spoken of about DiMaggio was—at least for this season—not a by-product of his reputation. He actually did it. [...]
Amazingly enough, sophisticated modern baseball math and the 1940s sportswriters are in alignment. In one of the most stories season in sports history, Joe DiMaggio was indeed the American League Most Valuable Player.
Kenny certainly makes it appear as though DiMaggio’s clutch moments push him past Williams. What I wish to argue is that context-dependent statistics are delicate weapons that must be used properly, and given this, a context-neutral approach should still be the optimal way to evaluate end-of-season awards like the MVP.
When I interviewed Kenny to talk about his book, he pointed out the generally equal status of a context-neutral and context-dependent approach. "I ask everybody now if they are a context-neutral guy or a context-dependent guy," Kenny told me. "Which are you? There's no right or wrong answer." As the aforequoted excerpts from Kenny’s book should demonstrate, however, Kenny does believe that the context-dependent answer is the right one for deciding the 1941 AL MVP race.
There’s no arguing that DiMaggio had the better context-dependent season, posting nearly 84 more runs than expected in 1941, relative to the runners on base and the outs recorded whenever DiMaggio stepped into the batter’s box. Williams, meanwhile, "only" created about 67 more runs than expected. Furthermore, DiMaggio posted a 7.27 Win Probability Added, compared to Williams’s 6.15. DiMaggio gave his team more chances to score, and more chances to win. Done and dusted. The Clipper is the champ.
Why is this enough? Kenny seems to think that since the Red Sox scored the most runs in the AL in 1941, Williams would have had plenty of chances to create runs and drive up his team’s win expectancy. After all, the Red Sox finished second in the junior circuit standings behind the Yankees that season. They were a good team. Williams had the opportunities. He just didn’t take them as often as DiMaggio.
There's just one big problem: Williams's clutch stats are incomplete. Retrosheet and Baseball Reference only have play-by-play data for 54% of games in 1941. While DiMaggio's are complete, Williams is missing play-by-play data for 68 of the 143 games he played that year. To compare the two players by win probability and run expectancy is utterly unfair.
Take May 7th, 1941 as an example. The Red Sox are at Comiskey Park playing the White Sox. Williams goes 3-for-4 with a walk and two home runs, driving in three of the Red Sox's four runs in an 11-inning, 4-3 victory. We don't have play-by-play data for that game, so Williams's run expectancy and win probability both read as 0.00. Even though Williams more or less single-handedly won the game for the BoSox, he got no credit for it in the eyes of WPA and run expectancy.
Now, it could very well be that DiMaggio would still come out on top in clutch stats were Williams's numbers complete. At this point in time, however, it can't be conclusively proven. Until we get play-by-play data for those missing 68 games of Williams's season, we'll never know.
It's not just that context-dependent stats are incomplete before 1974, and non-existent before 1930. Williams would still have been the superior AL MVP candidate, because a context-neutral approach is still more sound. That is because of a very simple reason: Players can’t control their clutchness.
While it’s true that all batting events add or subtract from clutch stats—whereas RBI are wholly dependent on guys being on base when you step up to the plate—the potential individual value of your plate appearances is going to change based on the base-out state. If you come up to bat a lot with guys on base, or if you frequently have a chance to drive in the game-winning run in the late innings, the potential value of your clutch stats will be a lot higher than if you were hitting in a lot of empty-base situations.
Let’s first consider DiMaggio. Yankees manager Joe McCarthy batted DiMaggio in the cleanup spot in his historic season. That should be no surprise, since DiMaggio was the Bombers’ best hitter. Right fielder Tommy Henrich batted third that year. Henrich had a typically strong season in ’41, posting a 137 wRC+ with a .377 on-base percentage. In other words, when DiMaggio got in the box, he would often find Henrich out on the basepaths.
Up in Boston, Williams batted third for 34-year-old Red Sox player-manager Joe Cronin. Cronin slotted in right fielder Lou Finney to hit in front of the Thumper. Finney was a classic light-hitting #2 hitter, managing a mere 91 wRC+ and a .340 on-base percentage in 1941. Finney got on base a fair amount, but compared to what Henrich did for DiMaggio, Williams just had fewer high run expectancy opportunities.
If anything, the lack of Williams’s high run expectancy chances relative to DiMaggio falls at the feet of Cronin, rather than Teddy Ballgame. Both Cronin and Jimmie Foxx had great seasons that year, each with plus-.400 OBPs and wRC+ scores in the 130s. Trouble was, both Cronin and Foxx hit behind Williams. If Williams smacked a double with the bases empty and nobody out, it was worth less (from a context-dependent perspective) than the same double hit by DiMaggio with Henrich standing on second. Williams couldn’t control that situation.
Therein lies the danger of evaluating players with context-dependent stats. We all know that WPA and RE24 aren’t predictive; there’s no skill involved that isn't captured in context-neutral stats. Time and time again, it’s been proven that hitters will hit the same no matter the inning, the score, the men on base or the outs recorded. While clutch hitting as a purely descriptive statistic has some value as a narrative tool, it does virtually nothing to tell us what control the hitter had over the event itself.
To his credit, Kenny concedes that WPA and RE24 have no predictive power, but he still insists on using them to judge players’ seasons against each other. If Ted Williams couldn’t control Joe Cronin’s lineup card, or that Tommy Henrich got on base more than Lou Finney, how can we truly say that Joe DiMaggio had a better 1941?
That’s what makes Kenny’s argument so strange. He uses language like "DiMaggio was better" when analyzing his run expectancy and win probability numbers. "Better" may seem innocuous in this context, but it actually invites the reader to infer something more deeply evaluative than what WPA and RE24 are actually telling us. The more precise way to understand DiMaggio’s higher clutch stats would be to say something like, "DiMaggio got hits in more run-scoring and win-deciding opportunities than Williams." It’s a little clunkier, but it decreases DiMaggio’s agency in an appropriate manner.
Kenny goes further, using these clutch numbers to confirm DiMaggio’s "game-winning ability" (emphasis added). WPA and RE24 tell us that DiMaggio got big hits in big situations to help his team score runs and win games, but they don’t tell us that it was due to any innate or developed skill on DiMaggio’s part. He was the same (admittedly great) hitter whether there were two outs with the bases empty in the top of the second and no score, or whether there were no outs with the bases loaded in the bottom of the ninth and the Yankees down by one. If he did change as a hitter in the big moments, neither WPA nor RE24 definitively prove that. They only tell us that he got hits when it counted. There was no "game-winning ability" that we can prove DiMaggio possessed in 1941.
The hitting streak that DiMaggio put up that summer in 1941 is practically mythical at this point. It doesn’t prove that DiMaggio was a better player than Williams, nor does it prove that DiMaggio should have been the AL MVP. Its spell is still cast. Its power remains to this day, even on social media.
The streak is a record that is nearly impossible to break. Add in the platitudes about DiMaggio’s hustle, playing the game hard every single day, doing all the little things well, and doing everything to help the team win, and it’s easy to be seduced by the magic of it all. Who could be faulted for enjoying such a narrative?
Ted Williams posted a 221 wRC+ in 1941, which is tied for the eighth-best hitting season in history. He walked nearly a quarter of the time he stepped to the plate, while also posting a .329 isolated power percentage. It’s incredibly difficult to argue that Williams didn’t do everything in his power to try and help his team win. Because his WPA and run expectancy numbers are incomplete, he may have actually done more to help his team win than DiMaggio. Since we don't know definitively, it's sloppy to argue for DiMaggio using clutch stats.
Furthermore, Williams also beat DiMaggio in fWAR by a win and in bWAR by a win-and-a-half. It's even more difficult to blame Williams for the talents of Lou Finney, or for the questionable lineup decisions made by Joe Cronin. The best data we have tell us that Ted Williams was the best player in the American League in 1941. The writers got it wrong; The Splendid Splinter should have won the MVP award that year.