I’m watching the New York/Atlanta game on ESPN tonight. The ESPN crew is doing a pretty good job—I like Orel Hershiser’s commentary—but they just made a big gaffe that few people will have caught. With the Mets leading 3-1 in the top of the eighth, the Braves loaded the bases with two out. To my delight, ESPN referred to the Mets’ win probability in that situation, stating that it was 90%. That struck me as a high figure, so I looked it up on my WPA spreadsheet and found that the Mets’ win probability at that point was actually 81%.
I don’t know where ESPN gets their figures, but they ought to verify them. There are many batting situations in which the sample size is just too small to rely on historical probabilities, and this was likely one of them. For instance, from 2000 to 2006, there have only been 58 games that reached that type of situation (run differential of -2 for the visiting team). The home team went on to win 88% of the time, but you can’t make general statements from just 58 instances. They should rely on the math.
Fittingly, Jeff Francoeur hit a single to tie the game. So much for your win probability.