Now that Nate Silver has gone to ESPN, one can wonder if SaberStats are going to become a bit more mainstream. This post looks at some current Giants players through the lens of WPA (win probability added).
The importance of situational hitting: Hitting with RISP, and more generally the need for a key hit or hits that makes the difference between winning and losing is a major theme and point of discussion in baseball. Who came through in the key situation, who did not? One can try to just remember, however, Bill James teaches that our memory is unreliable. One can also attempt to measure this by looking at batting average with runners in scoring position (RISP). This is useful, but somewhat limited. For retrospective analysis (looking at past performance) the best statistic in the sabermetric quiver is "win probability added" (WPA). WPA measure the contribution to winning (or losing) from every plate appearance and credits the batter accordingly. WPA recognizes the value of a single that drives a runner from 1st to 3rd in late innings in a tie game (a key hit, but one the involves no RISP and no RBI). WPA accurately measure the significant impact of a 2 run single in the 9th inning of a one run game, as well as the negative contribution of a double play ball. WPA can be positive or negative depending on how well the batter does. The average WPA of all major league batters is designed to be zero. (Baseball is a win-lose game.)
We can illustrate wih some examples. Buster Posey's game tying run home run in the 8th inning vs the Diamondbacks on April 22nd increased the Giants probability of winning by 40% (he gets a WPA credit of .40 wins for that); Belt's subsequent single in the 9th increased the Giant's probability of winning from 68% to 100% (for a credit of .32 wins). In another game, Blanco's ground-rule double in 8th inning vs the Padres (score 3-2) contributed 0.19 wins to Blanco's WPA. These are extreme examples with large positive WPA contributions. Mostly contributions, which can be positive or negative, tend to be much smaller. In general, WPA looks at every at-bat with unbiased eyes and records every result in terms of how it effects winning and losing. Overall WPA is a combination of many small positive or negative contributions, and a few larger ones. (Most contributions fall somewhere on a continuous spectrum between about -0.2 and 0.2.)
The next section looks a Giants starters (and Abreu) ordered by WPA on July 22, 2013.
Posey 2.96 wins. Posey leads the Giants in WPA. This is no surprise. Posey hits for high-average (.323), has power and a high on-base percentage. (His "line" (avg/on-base percentage/slugging) is .323/.393/.539. Posey's plate discipline is excellent. (He recognizes balls and strikes ver well and has a high walk to strike-out ratio.) Posey's hitting has also been timely. Posey's WPA is 2.96, which is 5th in the National League!
Belt: 1.49 wins. Though Belts WPA is only half of Posey's, it is the 2nd highest WPA on the Giants. Despite an average line (.267/.340/.449 (OPS of .789)) and a fairly high strike-out rate (23%), Belt's situational hitting has been well above average. That is, Belt has tended to hit well in key situations in close games. Belt is batting a modest .278 with RISP, barely above his overall batting average, however, in high-leverage situations* Belt has a batting average of .375. One could tend to dismiss this as being a "small sample size" result, and that may be the case, however, one should note also that Belt was 3rd on the Giants last year in WPA (behind only Posey and Cabrera). This is not to say that Belt is having a great year, but perhaps significantly better than one might think based merely on his conventional average or power stats. His WPA is 27th in the National League and 8th among NL 1st baseman.
Pagan: .96 wins. (46th in NL)) Despite playing only part of the season Pagan is 3rd in WPA. His inside the park home-run helped, and in general his hitting has been good and timely. Unfortunately he is on the DL.
Blanco: .64 wins. (51st in NL)) With a .270 average and, more importantly a modest .333 on-base percentage, Blanco's line is not overly impressive, however, he has had some hot streaks and some of his hitting had been very timely, thus he has contributed significantly to winning (at the plate). His average and on-base percentage are almost identical to Crawford's (see below). Blanco's WPA is quite respectable at 0.64 and his puts him 51st in the National League at this point in the season.
Scutaro: .55 wins. (62nd in NL) Scutaro is an all-around good hitter with an exceptional ability to recognize balls and strikes (plate discipline), an outstanding walk/strike out ratio of 1.39. Scutaro's WPA of .55 wins is quite good and places him 62nd in the National League. Scutaro has hit well in clutch/high-leverage situations throughout his time with the Giants. He continues to do so despite having to battle with injuries.
Sandoval: .35 wins. Pablo has had some great moments but has been troubled and sidelined by injuries. Overall, he is not having a great year. Pablo's strike-out rate is fairly low (14%), however, his on-base-percentage (.318) and his slugging (.401) are also quite low. His WPA, is probably negatively impacted by his high ground-into-double-play rate (Sandoval leads the Giants with 12 GDPs this year). In a good year, he makes up for that with great and timely hitting. I hope he can do that in the 2nd half.
Abreu: .23 wins. Abreu is included in this list because his 0.23 win WPA in only 69 plate appearances projects to roughly a 1.1 win WPA at 345 PAs, which would put him 3rd on the Giants behind Posey and Belt. Abreu has a .286 average. His hitting has been clutch and timely. Admittedly, this is a small sample size.
Crawford: -.35 wins. In mid May Brandon Crawford was 3rd on the Giants with a WPA of 1.18 wins. Since then Crawford has struggled. He is currently hitting .269, but only .171 with RISP. Hopefully he can get back to his early season form. A short-stop who can hit well is extremely valuable.
Pence: -.36 wins. After some strong early season numbers in terms of average, power and RBIs, Pence has dropped to an OPS of .759, 40 pts above Sandoval. However, despite an OPS well above Sandoval, his WPA is much lower. This (negative) WPA appears to be the result on not producing in key situations and some bad luck. From 2010 to 2012, Pence had WPAs of 2.51 wins, 3.12 wins and 2.0 wins, so this years partial result appears to be an aberration. Pence is an inspirational player who hustles on every play. I expect him to substantially improve his WPA before the end of the year. This expectation is justified by looking at Pence's non-situational stats which tend to predict a higher WPA for him than he has achieved so far.
WPA is a situational hitting stat available at a number of sites including fangraphs (a site was started by David Appelman about a decade ago). Here we focuses on the WPA for batters which is compiled by looking at the change in situation produced by each plate appearance a player has. If the batter's result (a single, a walk…) has improved his teams odds of winning, the player gets a positive credit . The size of that credit depends on the situation. If the batters result (a strike-out, a double-play…) has diminished his teams odds of winning, the player gets a negative credit (which also depends on the situation as well as what the batter did). Technically, the player is being compared to a hypothetical league-average player in each case. (There is also a WPA for pitchers.)
WPA is the most meaningful and accurate statistic for measuring what a player has done in the past. It is situational in the sense that the credit the batter receives for a single depends on the situation. If it is a game-tying single, the batter gets a larger credit; if the score is 10-1, the batter gets a smaller credit. Situational hitting is an important key to winning. Hence the emphasis on situational stats in understanding how a team acquired its won-lost record. While large swings in WPA are not possible, in practice WPA generally exhibits a gradual progression from game to game and is acquired through better than average performance over time. For example, Posey and Belt have been 1 and 2 for about 2 months. I think it accurately reflects what you would see if you rematch every plate appearance.
WPA is more predictive than batting average, but less predictive than on-base percentage (OBP). (Technically, looking at year-to-year correlation, WPA has an r^2 of about; batting average has an r^2 of about .15; OBP has an r^2 of about .33.) There are non-situational stats that are as predictive as OBP, such as wRAA (weighted runs above average), however, for looking at a players contributions to winning in the past, WPA is the best and most accurate stat. (WAR uses wRAA for its offensive component, making WAR more predictive, but less accurate in representing the past.)
* Batting average in high-leverage leverage situations is available at Fangraphs under "splits" for any particular player.