October 15, 2006
WPA Update: Season Totals
During the season, I posted four "WPA Updates" looking at where the Twins stood in a stat called Win Probability Added. Trying to explain WPA is both complicated and boring, plus I've already attempted to do so here in the past. Rather than give it another try, I'll instead recommend that anyone interested in learning more about WPA to decide whether or not it's a worthwhile tool should read this WPA primer from my Hardball Times colleague Dave Studeman or go digging around at FanGraphs.
I happen to think WPA is both fascinating and somewhat maddening, which is why I post updates, but no more often than every six weeks or so. It's an intriguing way to evaluate individual performance and often reveals worthwhile information, but also opens a can of worms that can get sort of messy. If you'd like to check the previous WPA updates, you can see how things looked through 18 games, 56 games, 86 games, and 123 games. It's a unique way to relive the season.
And now, here are the season-ending totals:
Joe Nathan 516 Terry Tiffee -1
Justin Morneau 446 Josh Rabe -8
Johan Santana 410 Brad Radke -14
Francisco Liriano 301 Jesse Crain -14
Michael Cuddyer 252 Mike Smith -25
Juan Rincon 242 Phil Nevin -34
Joe Mauer 236 Matt Garza -49
Dennys Reyes 147 Lew Ford -53
Pat Neshek 112 Jason Kubel -74
Jason Tyner 57 Torii Hunter -90
Boof Bonser 55 Tony Batista -95
Jason Bartlett 51 Luis Rodriguez -98
Matt Guerrier 34 Nick Punto -106
Ruben Sierra 27 Scott Baker -115
Shannon Stewart 22 Rondell White -116
Mike Redmond 21 Juan Castro -121
Glen Perkins 9 Luis Castillo -124
Willie Eyre 9 Kyle Lohse -147
Alexi Casilla 1 Carlos Silva -164
Chris Heintz 0
Perhaps the loudest objection raised with each WPA update is that little-used guys like Ruben Sierra, Glen Perkins, and Alexi Casilla boast higher totals than season-long contributors like Mike Redmond, Brad Radke, Nick Punto, Luis Castillo, and Torii Hunter. That's an easy way to dismiss the numbers if you're looking to do so and it's admittedly confusing at first glance, but it's important to realize that WPA isn't a traditional "counting stat" like RBIs or runs scored.
Sierra wasn't worth more than Redmond solely due to a 27-to-21 lead in WPA, because WPA actually shows how far above average someone has been. With "average" as the baseline, analyzing the totals goes beyond merely adding them up. "Average" is a lot more valuable than it sounds, because at the most basic level a roster full of players who are exactly "average" would go 81-81, the "average" hitter bats .275 with 20 homers, and the "average" pitcher has a 4.50 ERA. And they'd have zero WPA.
Sierra had 27 WPA in 33 plate appearances thanks to a couple key hits off the bench in his brief stint with the team. Meanwhile, Redmond had 21 WPA in 190 plate appearances thanks to batting .341 as Joe Mauer's season-long backup. Because of how difficult simply being "above average" is, much of Redmond's value actually stems from coming to the plate 190 times without dragging the team down. In other words, simply treading water in a pool of average has lots of value.
The lengthy list of everyday players with negative WPA totals shows that "average" production for long stretches isn't easy, which explains why Sierra's WPA total is far less valuable than Redmond's and ultimately insignificant. Similarly, Radke's -14 WPA doesn't mean he was a bad pitcher, but rather that he tossed 162.1 innings while more or less keeping the Twins on track for a .500 record. Perkins can claim a positive WPA total that's better than Radke's, but he can't come close to Radke's value.
Along with the above-average issue, further adjustments should be made for defense and positional differences offensively, neither of which are accounted for by WPA. A prime example is Juan Rincon's 242-to-236 lead over Mauer. Lost in Rincon's higher WPA total is that Mauer contributed a ton of value defensively, catching over 1,000 innings for the league's second-best pitching staff and gunning down 38 percent of base-stealers, while Rincon barely used his glove at all.
Beyond that, 236 WPA from a catcher is more valuable than 242 WPA from a reliever, because catcher is perhaps the most physically demanding, least offense-driven position in baseball. WPA is based on comparing players to "average" without actual positions factored in, but "average" for a catcher is much different than "average" for a middle reliever. In other words, whether it's comparing Sierra to Redmond or Rincon to Mauer, using WPA to evaluate value is about more than adding up totals.
Even with the above WPA totals as a jumping off point, it's relatively difficult to properly adjust for playing time, defense, and position without completely revamping the whole system. Instead, I've tinkered with the numbers in an effort to find a middle ground that gives weight to those issues, makes adjustments, and spits out a more complete, all-encompassing ranking. Rather than any overly complicated math, I focused a few simple, logical changes.
Relative non-factors like Chris Heintz or Mike Smith were ignored, while high-playing time regulars got a bump. Strong defensive players received a boost, with catchers, shortstops, second basemen, and center fielders getting an extra bonus. Position-based adjustments were applied to everyone, weighing each hitter's production against the average at his position and comparing each pitcher's production to the average player in his role.
Once I accounted for all of that, here's what my WPA-based MVP ballot looks like (I'll leave out the least-valuable guys, so as not to embarrass anyone who might be making $4.3 million next season):
1. Johan Santana 11. Pat Neshek
2. Joe Mauer 12. Boof Bonser
3. Justin Morneau 13. Torii Hunter
4. Joe Nathan 14. Brad Radke
5. Fran Liriano 15. Luis Castillo
6. Michael Cuddyer 16. Nick Punto
7. Juan Rincon 17. Jason Tyner
8. Jason Bartlett 18. Matt Guerrier
9. Dennys Reyes 19. Jesse Crain
10. Mike Redmond 20. Shannon Stewart
Not only did Johan Santana become just the eighth pitcher in baseball history to win the MLB Triple Crown by leading both leagues in wins, ERA, and strikeouts, and not only will he win his second Cy Young Award later this month, he also led all big-league starters in WPA:
Johan Santana 418
Roy Oswalt 415
Brandon Webb 369
Chris Carpenter 338
John Smoltz 328
Derek Lowe 323
Roy Halladay 320
Chris Young 314
You'll notice that places 2-6 are taken by NL hurlers. According to WPA, Santana was 31 percent better than his closest AL competitor, Roy Halladay. The value of an "average" performance from a starting pitcher is so high in terms of WPA that Santana is my clear choice for WPA-based team MVP. He was the best pitcher in baseball, led the league in both games started and innings pitched, and was able to post a 400-plus WPA when only two other starters topped even 350.
It may seem like a stretch for Mauer to jump ahead of Justin Morneau despite a 446-to-236 spread in WPA, but the difference in their defensive value alone makes a big chunk of that gap disappear. Further adjusting their offensive contributions for the fact that catchers hit far worse than first basemen makes Morneau's edge vanish. There are plenty of sluggers who matched or bettered Morneau's WPA total, but only Brian McCann can keep up with Mauer among catchers.
The rest of the ballot shouldn't be too surprising. Torii Hunter ranks lower here than he would on most fans' ballots, because WPA penalizes him for failing to come through in key spots (he hit .247/.319/.400 in "close and late" situations). Based on raw WPA totals, he actually ranked as the 10th-worst player on the team, but with defense and a positional adjustment offensively thrown into the mix, he moves way up. Somehow I doubt that'll change my reputation in some circles as a Hunter-basher.
Along with trying to determine who was most responsible for the Twins going 96-66, WPA also shows how top-heavy the team was. The Twins had five superstars (until Francisco Liriano went down), got star-level WPA from Michael Cuddyer and Rincon, and then saw a huge dropoff. They had three guys above 400 WPA and seven above 200, but just two from 75-200. For comparison, the Tigers had zero players above 400 and five above 200, but seven from 75-200.