Making Sense of the MVP Races

There’s quite a lot of bickering in sports, and not many things bring out more vehement disagreement than discussions involving who should get various awards. Even now, nearly 30 years later, when I think about Mo Vaughn beating out Albert Belle for the 1995 AL MVP, or Dante Bichette finishing second in that year’s NL race despite putting up just 1.8 WAR, I have to suppress a compelling desire to flip over a table. This year, thankfully, it’s hard to imagine the MVP voting results will be anywhere near as egregious as the ones we saw in ’95. That’s because the way MVP voters in the BBWAA evaluate players has changed dramatically since then.

Aaron Judge has easily the best traditional case for the AL MVP award if the season ended today. He leads the league in two of the main old-school batting stats: home runs and RBI. Bobby Witt Jr. and his .347 batting average is all that would stand between Judge and the Triple Crown. For what it’s worth, Judge would win the MLB Triple Crown, with twice the emeralds, rather than the AL one.

For most of baseball history, beginning with the first time the BBWAA handed out the award in 1931, numbers like these usually would’ve been good enough to win MVP honors. It also would’ve helped Judge’s case that the Yankees have one of the best records in baseball. If this were 30 years ago, Judge would all but officially have this thing wrapped up, barring an injury or the worst slump of his career.

But it’s the 2020s, not the 1990s, and I doubt anyone would dispute too strenuously the notion that ideas on performance, and their related awards, have shifted in recent years. Now, when talking about either an advanced offense statistic like wRC+ or a modern framework statistic like WAR, Judge certainly is no slouch. He currently leads baseball with 8.3 WAR, and his 218 wRC+ would be the eighth-highest seasonal mark in AL/NL history, behind only seasons by Barry Bonds, Babe Ruth, and Ted Williams. But by WAR, his lead is a small one, roughly two-tenths of a run (!) over Bobby Witt Jr., who has surged since the start of July (.439/.476/.803, 247 wRC+ in 33 games) to supplant Gunnar Henderson as Judge’s main competition for the award. Henderson was right there with Judge for much of the early part of the season, and though he’s fallen off a bit, he’s still fourth in the majors with 6.4 WAR and capable of catching fire again at any time. With a month and a half left, Juan Soto can’t be completely counted out either.

A similar dynamic persists in the NL. Shohei Ohtani has looked a lot like the obvious MVP choice for much of the season, as he’s done, well, one half of the Shohei Ohtani thing: He is murdering baseballs and pitchers’ dreams. But as with Judge, there’s some serious competition when you look at WAR. Ohtani stands at the top, but by a fraction of a run ahead of Elly De La Cruz. Ketel Marte and Francisco Lindor are both within five runs of Ohtani, and nobody serious has ever claimed you can use WAR to conclusively settle disputes on differences that small. De La Cruz has more WAR than Ohtani since the start of June, and the latter two have more than the Dodgers slugger since the beginning of May. Marcell Ozuna, who has strong traditional stats (.302 BA, 35 HR, 90 RBI) shouldn’t be completely discounted if the Braves show signs of life; those numbers still matter, just not to the extent that they once did. With a fairly wide open race, there are plenty of stars with name power lurking just behind the leaders, such as Bryce Harper and Freddie Freeman.

The answer of who should win the MVP awards is one we probably can’t answer beyond me giving my opinion, which I won’t do given the likelihood that I will be voting for one of the awards. But who will win the MVP awards is something we can make a reasonable stab at predicting. It’s actually been a while since I approached the topic, but I’ve long had a model derived from history to project the major year-end awards given out by the BBWAA. It was due for some updates, because the voters have changed. Some of the traditional things that voters prioritized, like team quality, have been de-emphasized by voters, though not completely. And the biggest change is the existence of WAR. Whatever flavor you prefer, be it Baseball Reference, Baseball Prospectus, or the smooth, creamy swirl that can be scooped by our display window, this general stat has changed a lot about how performance is perceived.

There have been 47 MVP awards presented to position players who finished their seasons with fewer than 6.0 WAR; that’s more than a quarter of all hitter MVP seasons. However, excluding 2020, a hitter has not won an MVP without reaching that threshold since ’06, when both winners fell short: the NL’s Ryan Howard had 5.92 WAR, while AL winner Justin Morneau had 3.77 WAR.

When modeling the data, I use all the votes, not just the winners, and WAR is a pretty lousy variable when predicting voter behavior throughout most of history. That’s not surprising on its face since we’ve had WAR to use for only the last 15 years or so, making it impossible for most awards to have explicitly considered it. But there also appears to be only marginal implicit consideration, in which voters based their votes on the things that go into WAR without using the actual statistic. There’s a great deal of correlation between winning awards and high WARs in history, but that’s only because two of the things that voters have really liked, home runs and batting average, also tend to lead to higher WAR numbers. As an independent variable, WAR doesn’t help explain votes very well. That is, until about the year 2000.

If you only look at votes since 2000, all of a sudden, WAR goes from an irrelevant variable to one of the key components in a voting model. Voters in 2002 may not have been able to actually look at WAR, but even before Moneyball was a thing, baseball writers were paying much more attention to OBP, SLG, and defensive value at least partially because of analysts like Bill James, Pete Palmer, and John Thorn in the 1980s and ’90s. Now, depending on your approach, once you deal with the correlations between variables, WAR comes out as one of or the most crucial MVP variable today. Could you imagine a world, even just 20 years ago, in which owners would propose paying players based on what sabermetrics nerds on the internet concocted?

The model I use, which I spent most of last week updating, takes modern voting behaviors into consideration. I use all three WAR variants listed above because it’s not clear which one most voters use. Here is how ZiPS currently sees the two MVP races this season:

This model thinks Judge is the favorite, but his odds to lose are nearly a coin flip. Witt is the runner-up, followed by Soto, Henderson, and the somehow-still-underrated José Ramírez. If we look at a model that considers all the BBWAA-voting years rather than just the 21st century results, this becomes a much more lopsided race.

Over in the NL, the updated ZiPS model sees a race that’s far more uncertain than the one in the AL.

Ohtani comes out as the favorite, but he has less than a one-in-three chance to win it. Behind him are the other WAR leaders, plus Ozuna.

Some of the WAR leaders without strong Triple Crown numbers, like Lindor, drop off considerably based on the entire history of voting, while Ozuna becomes a co-favorite with Ohtani. I haven’t talked about pitchers much in this article; they’re still included in the model, but none make the top 10 in the projected probabilities. Simply put, the willingness to vote pitchers for MVP seems to have declined over time. ZiPS doesn’t think any pitcher has been as dominant this season as the two most recent starters to win the award, Clayton Kershaw in 2014 and Justin Verlander in ’11, and closers these days typically can’t expect to get more than a few stray votes at the bottom of ballots.

It’ll be interesting to see how voting continues to change moving forward. In any case, no matter who you support for the MVP awards, strap in because there’s still plenty of baseball left to be played.