from Amazon
Monday, April 09, 2007
WPA to ARod: I love you
Here’s what that great game feels like in graph form. (I also put a comment on that thread that some here might find interesting.) What is also interesting is how ARod looks so far this year: negative clutch!
How is this possible? Well…
A little quirk in the way David calculates clutch (which is felt in the early seasons), plus a WPA quirk.
You see, the right way to do it is to take WPA one PA at a time, and divide that by the LI of the situation. This gives you the player’s WPA, without the extra impact of the game state. And you compare that to his Linear Weights for that PA. In A-Rod’s case, Yankees were down by 1 in the bottom of the 9th, with bases loaded and two outs. This is the highest possible leverage in a game: 10.9.
And what happened? Well, a standard HR adds 1.40 runs. A standard runs to win converter is around 11 or so, meaning a HR regardless of game state would add +.13 wins. But, in ARod’s case, it added +.72 wins, or about 6x more than it normally would.
You have here a problem: the PA here impacts the game 10.9 times more than normal, but a HR impacts the game just 6 times more than normal. This HR should have been worth .13 x 10.9 = 1.4 wins (an impossibility).
Anyway, getting back to David, he looks at his season-to-date LI and sees a 1.90, meaning that, on average, ARod’s PA are double the importance than a random PA. This is the effect of that 10.9. But, the problem is that David expects each of his PAs to have a 1.90 times effect. So, ARod so far is +0.52 in OPSwins, but only +0.84 in WPA. But, with double the LI of normal, we expect his WPA to be +1.04 if there was no clutch. That difference is .20 runs, of chokiness!
How to correct it? Look at each PA one at a time. ARod is +.52 OPS wins in 24 PA, or an average of +.02 OPS wins per PA (which itself is fantastically high… a great hitter would be +.01 wins per PA). In his super-clutch PA, he was +.72 wins, which after we divide by 10.9, brings us down to +.07 wins, irrespective of game state. The difference between .07 and .02 is .05, and that’s the extra impact of ARod’s clutch, without respect to the game state. Or, you can look at it the other way: he’s normally +.02 this year, which if you multiply by 10.9 means that we expected him to add +.22 wins based on his seasonal data. He add +.72 wins, meaning he added +.50 wins to his team over and above his 2007 production.
Just some little quirks to keep you thinking…