Every year, we debate who’s going to be the MVP. Some people use their eyes and their gut … others use WAR or another metric to be more objective about their feelings.
I have been doing a lot of thinking/writing about Win Probability Added (WPA) and thought of another way to measure value. Who has increased their team’s chances to win more than they might be expected to?
If we define a typical player’s contribution to their team as 1/9 of the offensive roster, then we search for the players who have the most games with a WPA over (1/9) or 0.111 (since WPA is calculated out to three digits). In word form, this means that each player is expected to contribute 11.1% of a team’s win, offensively. Since baseball-reference only does greater-than-or-equal-to, we search for # of games where a player’s WPA >= 0.112.
I think that’s a reasonable way to measure “value”, don’t you? It’s not perfect but it’s a decent starting point.
If we do it this way we get the following top ten in the AL:
- Chris Davis – 43 games
- Miguel Cabrera – 38 games
- Mike Trout – 32 games
- Josh Donaldson – 32 games
- Kendrys Morales – 30 games
- Josh Hamilton – 30 games
- Prince Fielder – 30 games
- Robinson Cano – 30 games
- Jason Kipnis – 29 games
- Edwin Encarnacion – 28 games
So by this measure, Chris Davis comes out ahead as the AL MVP; he contributed more than his share (at the plate, anyway) more often than anyone else in the AL. As an O’s fan, this makes me happy 🙂 I seem to recall that the O’s did most of their scoring late in games this year; WPA likes late scoring, so it’s somewhat less surprising to see Davis atop this list.
It’s also surprising to see Josh Hamilton so high. He was written off after his early-season struggles. I suspect that when he did hit, he hit in the clutch, therefore adding to his WPA contributions.
In the NL, we get:
- Paul Goldschmidt – 41 games
- Freddie Freeman – 34 games
- Andrew McCutchen – 33 games
- Hunter Pence – 32 games
- Ryan Zimmerman – 29 games
- Jay Bruce – 29 games
- Yadier Molina – 28 games
- Matt Holliday – 28 games
- Pedro Alvarez – 28 games
- Joey Votto – 27 games
The thing I like about this method is that WPA takes context (score, # of runners on base, etc) into account. To me, contributing in these situations is a big component of “value”. Even useful metrics like WAR don’t do this. There are reasons for that, mostly because WAR is supposed to measure a player’s skill, and getting-to-hit-in-important-situations is not a skill (I would imagine it’s some combination of batting position and luck).
But this metric has its flaws. For one, it doesn’t take either defense or baserunning into account. I believe these are a huge part of players’ games. So if a player commits an error, they aren’t penalized WPA. Also, this method is somewhat lazy since its starting assumption is that each player will add 1/9th of the team’s probability, offensively, to the team’s win. This isn’t true; off the top of my head I’d say we should adjust for a) the number of times a player batted relative to the others (maybe excluding pinch hitters) and b) that player’s overall talent level. After all, Davis is more likely to increase the O’s chances to win than, say, Steve Pearce simply because Davis is a better hitter.
To calculate the former, we could maybe try to predict the average number of PA a player is likely to see in a game, based on their lineup position and the average number of PA that lineup position sees in a game, and adjust accordingly. This wouldn’t work for players that shift around a lot, but Davis batted 5th for nearly 2/3 of the 2013 season, so we could just stick him there.
To calculate the latter, we could (just brainstorming here) take that player’s wRC+ to date in the season for each game they played in, or use a two-or-three-season lookback wRC+ average as a placeholder.
Of course we already have the metric WPA/LI, which puts players on equal footing by neutralizing the differences in the numbers and kinds of situations in which they come to the plate. By this metric, Davis drops to third in the AL (behind Trout and Cabrera), Cano moves up to 4th, Encarnacion rises to 5th, Donaldson drops to 6th, and Adrian Beltre enters the conversation at 7th place.
Finally, another problem with this method is that it ignores pitching. I don’t think that, for a single game, 9 batters will ever add up to 100% of a team’s win probability. If that’s the case, then the pitcher is giving up runs left and right! A good pitching performance will contribute more to a team’s win and thus the “expected” contribution of the offense is lowered.
Oh well – it’s still fun to think about, and maybe I will look into accounting for more of these discrepancies in the future.