I don't know much from baseball, but I would be most interested in the game situation cluster. The other stuff, I'm not sure if I would use any of it. The idea is that I want a purely actuarial account based on a large sample of data. I then adjust for "this particular game" factors based on what I'm seeing/other scouting. The latter is hard to capture statistically. My guess is that when the price deviates too much from what the actuarial data suggest, it's a good play. Also, odds information is stuff you usually had, so absent a complicated theory about how it interacts with game situations, it doesn't do much but complicate the model.
|