POSTED BY Coley Ward ON 11:02 am, June 26, 2007 - POSTED IN Rant
Those daredevils over at ESPN tested out a new statistic during this weekend’s College World Series — win probability.
It’s actually not a new stat. It’s just new to baseball. Usually, you’ll see the win probability stat used in poker.
How does win probability work? In poker, it’s easy. There are 52 cards in a deck, so after the players receive their first two cards a computer (or a really smart MIT grad) calculates the odds of each player winning the hand.
In baseball, it’s a little more complicated. Or, at least, it must be. Actually, I’m not entirely sure how win probability works in baseball. Apparently, Orel Hersheiser explained it the other night, but I missed his explanation. I’m told it has something to do with how each team has fared against opponents with similar records in similar spots. Or something like that.
My problem with win probability, as applied to baseball, is that it doesn’t really tell me anything. In poker, win probability is an exact science. There are only so many cards in the deck and only so many combinations of cards will allow me to beat my opponents.
But in baseball, it’s a lot more wishy washy. There’s no way to say exactly how likely or unlikely it is that a team will come from behind to win. No way to factor in the blister that Josh Beckett has on his finger. No way to calculate that Ryan Howard is in the zone, or that Derek Jeter has got a case of the dropsies.
As a result, all the win probably stat does is assign a number to the notion that it’s unlikely that a team trailing 5-2 in the seventh inning will come back to win the game.
But we already knew that, didn’t we? We didn’t need the number.
JojoFireball 
Paul Moro 
One recent study of baseball numbers I’ve read actually breaks down each element of the game into a “run value”. It has compiled data over the past ten years and calculated how much value each “act” (such as stolen base, ground out, double) has (or rather, is expected to have) on the scoreboard.
The study also looks at “Win Expectancy”, which seems like it’s close to what ESPN has done. “Win Expectancy” takes two teams of fairly equal talent and breaks down every possible situation in a game (2 down, 1 on, top of the fifth, and so on) and calculates how likey either team is to win from this point forth.
But as you point out, Win Expectancy pretty much states the obvious. But I do find it interesting to see how likely or unlikely comebacks are.