FanPost

WPA vs. WAR: The good, the bad, and the shortstops

I've been thinking a lot about WPA lately, and how it can be interpreted. It's a situational stat, but it comes to mind for me whenever somebody mentions a generally fine player being terrible after, say, grounding into a double play, or striking out with runners in scoring position, etc. So with that in mind, I went back through all the Brewers' first-half games to look at aggregate and per-game WPA for each player, as well as comparing it to WAR (Cheeseandcorn has a look at game-by-game WPA highs and lows here). Tables and numbers after the jump.

A couple things to note, first. WPA privileges batters in a way that WAR doesn't, as far as I know. Batters have a floor to keep them from taking too much negative WPA on any given plate appearance, while pitchers have a ceiling that tends to keep them from racking up big positive numbers. Relievers and starters are different because relievers are pitching later in the game, with more significance attached to every event in a close game.

That said, Aramis Ramirez leads the team in total WPA, with 2.988 over 78 games played. Granted, about a third of that came in one game earlier this month. But this isn't surprising for a clean-up hitter -- he's tied for second in the NL in doubles, and there tend to be runners on base when he hits them. That allows him to both drive in other runners and get a guy (himself) into scoring position. Ryan Braun is second on the team with 2.377 over 80 games. They're also #1 and #2 among batters in WPA/G. After them, it gets a little surprising (rates as percentages for ease of interpretability, plus WAR/G for comparison):

Batter Games WPA/G (%) WAR/G (%)
Ramirez 78 3.83 1.79
Braun 80 2.97 4.38
Gamel 21 2.11 -0.48
Lucroy 43 1.88 4.88
Maldonado 37 1.37 1.62
Kottaras 53 1.10 1.51
Hart 83 0.89 1.33
Ishikawa 42 0.40 -0.48
Gonzalez 24 0.40 -0.42
Gomez 62 0.30 0.00
Aoki 77 0.18 1.95
Weeks 81 -0.04 -1.98
Ransom 37 -1.27 1.35
Maysonet 30 -1.28 1.33
Green 44 -1.40 -0.45
Morgan 74 -1.67 0.54
Conrad 25 -3.46 -2.80
Izturis 43 -4.04 -2.33

These two are pretty well correlated (r = .64, p < .001 across the whole team), but the guys that are really divergent are what's interesting to me here. Look at Gamel and Morgan, in particular. Gamel was having a great start by WPA, and a totally mediocre one by WAR (-0.1 total for his truncated season). A lot of folks seemed pretty excited by Gamel's start, and that could easily have been fueled by strong situational hitting. Morgan's just the opposite, with his OK 0.4 WAR for the season, but his terrible WPA seems to justify the perception that he's retired Tony Clutch. Aoki's similar differential could be explained by how often he leads off (or hits #2 behind a guy who strikes out); with no one on base, his potential WPA is decreased.

Meanwhile, the shortstop situation is the clear black hole for the offense. Maysonet's numbers across a smallish sample are probably inflated a bit by his grand slam, but with him back in Nashville it's kind of irrelevant now (as are Conrad's awful numbers). The clear difference between Ransom and Izturis (who really hurts his WPA by grounding into double plays all the time) makes it extra irritating that Izturis keeps getting regular starts.

One other note: WPA/G is correlated with games played (r = .42, p < .01), meaning higher WPA/G is associated with more games played, but WAR/G is not (r = .12, p = .5). This tells me that RRR may be affected by WPA-maximizing play the same way fans may be.

And the pitchers (everybody with less than eight appearances* is excluded, and starters' brief relief appearances and batting numbers are included):

Reliever Games WPA/G (%) WAR/G (%) Starter Games WPA/G (%) WAR/G (%)
Veras 40 1.30 -0.75 Fiers 8 6.45 20.00
Loe 37 -0.01 -0.81 Greinke 19 3.54 13.16
Dillard 34 -0.44 -0.29 Gallardo 18 -0.49 7.22
Rodriguez 42 -0.79 -0.24 Marcum 14 -1.03 10.00
Parra 36 -1.52 -0.56 Estrada 13 -5.28 2.31
Axford 36 -3.79 -2.50 Wolf 19 -13.32 -6.32
Perez 10 -8.61 0.00

No point comparing them to the batters, obviously, but comparisons can be made among the groups. Not really a surprise that nobody's got positive WAR in the bullpen, but Veras's occasional holds and long scoreless (but not tortureless) stretch are apparently real things. Loe, too, looks kind of OK considering the ceiling that WPA places on pitchers.

Among the starters, you can really see that ceiling in effect. You can also see that the rotation had a pretty good first half, apart from Estrada's fairly frequent WPA-killing One Bad Innings, and also Wolf in general. His aggregate WPA (-2.530) is worst on the team, way behind Izturis (-1.739), and his total WAR (-1.2) is worse than everyone but Weeks (-1.6).

And then there's Fiers -- tops on the team in WPA/G and WAR/G, and those would be even higher without his inconsequential one inning of relief pitching. Amidst the general disappointment of the season, Fiers is a huge bright spot that's probably still being overlooked a bit. Greinke is #3 on the team in WPA/G (behind Ramirez, ahead of Braun), and he's generally made the idea that Gallardo was the #1 starter in April look a little silly. He and Fiers have both put up a lot of scoreless innings in close games, and as much as it's clear the Brewers could look much better with a different bullpen, they could also look a lot worse with a couple different starters.

* Excluded: Chulk (7 appearances), Hernandez (6), McClendon (4), Narveson (2), Peralta (1), Thornburg (1)