"A New York Treasure" --Village Voice

Color by Numbers: When the Going Gets Tough…

Over the last few weeks, a spate of losing has sent the Yankee Universe spiraling into a state of depression. So, what better time to take a look at a statistic called WPA?

Although many new sabermetric tools can be intimidating, Win Probability Added, or WPA, is a fun stat. Basically, it tries to determine the impact of every play on the outcome of an individual game. Instead of getting too caught up in the derivation, it’s better to simply think of WPA as a metric that identifies the pivotal points that often go unnoticed in a regular box score. In a sense, it is the mathematical equivalent of the late Bill Gallo’s classic “Hero and Goat” cartoons that appeared in the New York Daily News for over 50 World Series.

Unlike most context neutral statistics, which are generally favored in the sabermetric world, WPA rewards (or punishes) a player as much for the contributions of his teammates as his own performance. It also shifts the focus from what a player does to when he does it. As a result, it’s most useful for measuring the impact of a particular play, not the value of an individual player. At the risk of opening a whole new can of worms, WPA can also be considered a measurement of clutch, provided you define the concept in terms of the event that transpired and not the person who performed it (i.e., a clutch hit doesn’t necessarily imply a clutch hitter).

Last night’s 15th inning tug-of-war between the Yankees and Orioles is the perfect game to illustrate the usefulness of WPA. As evident in the win expectancy chart below, just about every play in a close game can have a significant impact on a team’s chances of winning. However, certain moments always emerge as turning points, and that’s what WPA helps to identify.

Yankees vs. Orioles, May 18: Win Expectancy Chart (click to enlarge)

Note: Probability is for winning team (Yankees).
Source: Baseball-reference.com

By taking these game-by-game snapshots and compiling them into a cumulative total, WPA highlights the players who have most often come through in the clutch. To no surprise, Curtis Granderson has boosted the Yankees fortunes more than any other batter*. His 1.4 WPA is not only almost three times the next closest contributor, but the total is good for sixth best among all American League hitters (although less than half of Jose Bautista’s league-leading 3.0). On the other end of the spectrum, several key Yankees have underwhelmed in terms of WPA. In particular, Brett Gardner, Nick Swisher and Derek Jeter have all posted a significantly negative contribution to win probability, which helps explain why the Yankees’ offense has been sputtering for so long.

*WPA is also calculated for pitchers, but we’ll save them for another post.

Yankees WPA Scores (Offense): Cumulative and by Outcome

Player Total Wins Losses
Curtis Granderson 1.381 1.242 0.139
Eric Chavez 0.483 0.587 -0.104
Russell Martin 0.468 0.636 -0.168
Mark Teixeira 0.414 1.064 -0.650
Robinson Cano 0.352 0.763 -0.411
Francisco Cervelli -0.001 -0.082 0.081
Jorge Posada -0.005 0.098 -0.103
Gustavo Molina -0.066 -0.039 -0.027
Chris Dickerson -0.082 -0.082 NA
Eduardo Nunez -0.099 -0.255 0.156
Alex Rodriguez -0.192 0.874 -1.066
Andruw Jones -0.331 -0.018 -0.313
Nick Swisher -0.701 -0.252 -0.449
Derek Jeter -0.921 0.190 -1.111
Brett Gardner -1.083 -0.211 -0.872

Note: As of May 18, 2011
Source: Baseball-reference.com

Looking at cumulative WPA can sometimes be misleading because several smaller contributions can be overshadowed by one epic result. In order to get a better idea of who has been “coming through” on a more consistent basis, it’s often better to look at WPA scores for individual games instead of one combined total.

Brett Gardner’s and Curtis Granderson’s Daily WPA

Note: As of May 18, 2011
Source: Baseball-reference.com

As evident from the graph above, the Yankees’ leader and laggard in terms of WPA have both earned their status over the course of the entire season. Granderson’s consistent contributions to winning are noted by the dominant splashes of blue above the x-axis, while Gardner’s frequent failures at key moments are outlined in red.

Finally, returning to concept of goats and heroes, the table below lists the number of times each player has had either the highest WPA in a win or the lowest WPA in a loss. Although rudimentary in terms of analysis, these classifications usually match closely with observation and sometimes help explain why a particular player has earned a positive or negative reputation when it comes to pressure situations.

Heroes and Goats: Players with the Highest/Lowest WPA in Wins/Losses

Source: Baseball-reference.com

So, how does WPA jive with your observations?

Categories:  1: Featured  Baseball  Yankees

Tags:  color by numbers  william j

Share: Share on Facebook Share on Twitter Share via email %PRINT_TEXT


1 Diane Firstman   ~  May 19, 2011 12:33 pm

Wow ..... very nicely presented!

2 ms october   ~  May 19, 2011 2:16 pm

really good stuff william.
i especially like the picture referencing the other wpa :}

i'm not sure how much to really take from the hero/goat ledger, but it was interesting nonetheless.

interesting that only 4 regulars have a positive wpa now. i guess that's why they are basically .500 right now (though this is removing pitchers from the equation).

as for the question - i thought teix would have a little lower wpa and arod a bit higher but otherwise it pretty much jives with my take.

3 Jon DeRosa   ~  May 19, 2011 2:25 pm

I like how you suggest to look at WPA for plays instead of players.

I am under the impression that WPA does not take into account the quality of the players in the situations it's calculating, correct? Like if the Cubs load the bases to pitch to Nick Punto it will add the same amount of "Win Probabilty" to the Cards as if the Cubs loaded the bases before Albert Pujols?

4 William J.   ~  May 19, 2011 2:31 pm

[2] The Hero/Goat ledger is strictly about fun. Obviously, if the same guy finished second in the hero category, he'd gain no recognition. It's kind of like the Hockey stars system.

[3]Your assumption is correct...no extra credit is given for quality of opponent. The same is also true for all the big context neutral stats, although I wouldn't be surprised if someone is working on that as we type.

5 The Hawk   ~  May 19, 2011 6:42 pm

Well done sir! though you want to say "jibe" there, not "jive"

6 William J.   ~  May 19, 2011 7:00 pm

[5] Thanks for translating, but until one of the Banter copy editors can fix it, I'll pretend I meant to talk a little jive!

feed Share on Facebook Share on Twitter Share via email
"This ain't football. We do this every day."
--Earl Weaver