from Amazon
Thursday, January 25, 2007
The Two Sides of Win Expectancy
An excellent dialogue between stat-heavy Dan Fox and Sterger-heavy Will Carroll. The game in question is also represented at Fangraphs, and it includes the more-important Leverage Index.
Without question, what both sides accept, or likely accept, is that this is a reasonable representation, in real-time, as the events unfold, of the likely chance of each team winning, as well as the perceived (or should have been perceived) pressure by all players that day. We should all be on board here.
What is in question is:
How to credit the change in the win probability. Will Carroll rightfully points out that the credit being handed out is, mostly, out of the hands of the players, when the leverage is so high. A guy gets credit for the outcome of his plate apperance double, triple, or even five times more, than a guy who came up just a few batters earlier. In these isolated cases, the players gets the huge plus or huge minus because of timing, because of where he is in the sequence.
On one side, he’s right. On the other side, so what? I’m pretty sure I’m smarter than a monkey not named Marcel, and yet ask a monkey to throw a dart on the board to pick ten stocks, and ask me to analyze the ten best stocks, and the monkey might end up beating me. And, he really did nothing other than be the one holding the dart. Yet, he’s the one who will potentially have a million dollars and partying in Vegas, while I am the one who is writing a blog about it.
Win Probability Added (WPA) simply gives out real cash, based strictly on who was there when the money was being handed out. Whether it required tons of talent, or the teeniest tinest of talent, to generate that cash is irrelevant to WPA.
However, given the chance to repeat this 100 times, I’m pretty sure I’ll end up being more successful than a monkey. I’m still writing a blog, but there’s no monkey writing Shakespeare. Yet.
This isn’t discussed in the article, I don’t believe, but I have just recently begun to look at the WinEx chart with the intent of judging the efficiency of sacrifice bunts.
A big factor is the quality of the batter; while WinEx chart assumes an average hitter, most sacrifices are asked of below-average hitters.
Is there a shortcut to adjust for this, e.g., if there is a situation where the WinEx is .717 with an average hitter, and I have a hitter who is, say, -20 runs against average per 150 games, is there an easy to way figure out what the WinEx is with this specific guy at the plate? Is is .712? .716? .657?
(I know that, ideally, you’d want to include platoon effects, as well as with groundball rates for the batter and pitcher, and break down performance by components and so forth; I’m just looking for a ballpark.)
I also will look at the chapter on this in The Book, which I don’t have in front of me at the moment.