WAR! What is it good for?

WAR! So you want to know what it’s good for?

The Temptations (and the more famous Edwin Starr) would have you believe that it is good for “absolutely nothing.” Baseball fans know that it is good for measuring value above a certain mark, the “replacement level,” so to speak. What is the replacement level?

Replacement level

Replacement level is defined as a baseball team made up entirely of such players with a winning percentage of .294 or a season of 48 wins in 162 games. Perhaps not coincidentally, a 48-win season is 33 wins under .500 and there are 30 major league teams. This means that if you add up all the WAR in the major leagues, you get 990 (let’s call it 1,000 for simplicity). There are 1,000 bWAR each season.

WAR might seem a bit too complicated for a normal person to calculate. I’m only smart enough to evaluate players and maybe introduce a coefficient to put them in some sort of order. In the NHL, statisticians have used “point shares” to divide credit (and blame) and pass it on to the various on-ice personnel. In the NBA, there are “win shares” which provide pretty much the same information (though I think PER is probably a better way to evaluate players, multiplied by minutes played, of course). In the NFL, there’s the much more esoteric “approximate value”, the basic methodology of which I don’t even understand. It’s a good thing I’ve sworn off football.

What WS, PS and AV have in common that WAR does not is that the first three are measured above zero and WAR is measured above an imaginary waterline. One could argue that the waterline in question is at a somewhat arbitrary point compared to the other three sports. So what would a “Wins Above Zero” calculation entail?

The simplest solution from a purely layman’s perspective would be to introduce a coefficient of 2.43. Multiplying everyone’s existing WAR by 2.43 gives a total WAR of 2430, which, not coincidentally, is also the number of games played per season. The problem with this is that anyone below 0.0 WAR has an even more negative WAR. In fact, a player with -0.5 WAR might have a WAR of around 3.0, but if you multiply -0.5 by 2.43, you get -1.215. That’s not an exact number. Only the absolute worst of the worst would have a WAZ below zero. Someone with Jose Abreu’s 2024 stats, for example, would likely have a WAZ close to 1.

What’s the methodology? Before I go any further with that, I need a real-world example of how WAR actually relates to a team’s W/L record. It’s never 100 percent accurate. For my example, I’ll use the 2023 Astros.

Houston won 90 games in the 2023 regular season, which means we should get 42 WAR if we add them all up. We actually get 46.5 bWAR, which is kind of in the range, but not close enough to derive the “perfect” value. To continue riding this wave, we need to understand that hypothetically this team would also finish with 90 WAZ. Multiplying by 2.43 gives us a total of 112,023, which is way off the mark. We need to approach it differently.

I’m not at the point where I can develop a formula to do the math from scratch in real time, but I can put the 2023 Astros at 90 and then reverse engineer how I got to that number. Maybe if we figure that out, we can develop a formula to monitor that domestic statistic on a daily basis. Could the answer lie in WPA? I bet it could.

Probability of winning added

Wins Probability Added is a situational statistic that measures impact in addition to base stats. A grand slam might count for almost nothing in a 15-4 win, but a game-winning RBI single with two outs could be in the 0.8 range or more. That said, when calculating Wins Above Zero, perhaps we should recognize those players who consistently hold on in crucial moments. Consider that a team with 81 wins has a collective WPA of 0.00 despite having a WAZ of 81. A team with 100 wins has 19 WPA and a WAZ of 100. A team with 48 wins despite having 48 WAZ (and 0.0 WAR) would have a WPA of -33.

An unintended byproduct of this new rating system will discredit those stat hounds who only come into play when a game is all but decided. Giancarlo Stanton springs to mind. When I followed the Marlins, it seemed like his best at-bats were when the game was already out of reach. So 0.0 WPA is worth 81 WAZ. Our 2023 Astros, the 90-win team, should have 9 WPA, and they do.

Next, a ton of research. I found no correlation between WPA and bWAR. For example, Jeremy Peña had a team-fourth 3.8 bWAR but had the worst timing for his good moves and performed best when it didn’t matter, with a team-worst -3.23 WPA. Alex Bregman, tied for the team-leading 724 batting appearances, had a team-second 4.9 bWAR but was a distant eighth on the team with 1.65 WPA. Additionally, Ryan Pressly had 0.1 bWAR and 0.01 WPA with 268 batters, but Cesar Salazar had 0.2 bWAR and -0.29 WPA with only 19 batting appearances. This has led me to the conclusion that I also need to factor in a player’s actual playing time somewhere in the calculation.

Playing time

You can’t underestimate how important it is for a team to be able to count on a player who stays healthy. Framber Valdez, for example, is the closest thing we have to a tough warhorse – and even he spent a few weeks on the injured list. Jose Altuve, normally a very tough player, had a few issues last season that resulted in him only playing in 90 games.

I don’t have a ready-made solution. All I know is that there has to be a formula to take existing WAR, WPA and playing times and put them into a coherent design, add all the results up to 90, and make sure none of the final results are very far below zero.

After a lot of trial and error, I came up with the following formula: (bWAR*.9677)+(PT*.002912)+WPA. If I plug in all these numbers, I get 90 wins. Here are the biggest winners (and losers) of the 2023 season.

Heroes and Zeroes (Wins Above Zero, entire team)

Kyle Tucker 11.85
Jordan Alvarez 10.62
Alex Bregman 8.50
Chas McCormick 8.15
Framber Valdez 7.48
Hector Neris 7.13
Jose Altuve 5.97
Bryan Abreu 5.23
Mauricio Dubon 4.07
JP France 3.84
Yainer Diaz 3.36
Cristian Javier 2.84
Justin Verlander 2.60
Jeremy Peña 2.29
Jake Meyers 1.53
Brandon Bielak 1.21
Phil Maton 1.19
Luis Garcia 0.99
Ryan Pressley 0.89
Hunter brown 0.87
Ryne Stanek 0.83
Jose Urquidy 0.64
Jose Abreu 0.57
Seth Martinez 0.56
Ronel Blanco 0.51
Kendall Graveman 0.47
Bennett Sousa 0.33
Matt Gage 0.28
Corey Julks 0.27
Michael Brantley 0.20
Shawn Dubin 0.07
Parker Mushinski 0.04
Grae Kessinger -0.02
Cesar Salazar -0.04
Joel Kuehnel -0.06
Ryan Bannon -0.36
Bligh Madris -0.48
Jon Singleton -0.93
Rafael Montero -1.03
David Hensley -1.09
Martin Maldonado -1.39

I think these results are pretty solid all things considered. Check back tomorrow to see if this formula can be directly applied to the 2024 team.