You are very much on the right track.
I also feel that a regression is unreliable way to introduce team strength.
I am trying to think what else I can say without giving away my methods for doing this, which I don't want to do just yet.
This might end up being my last season trading, as I plan to take a long vacation/retirement starting in early December. I have not yet decided what to do with my knowledge regarding interactives. Do I give it to no one? Do I give it to a select group of people? Do I give it to WSEX (if they want it), whose treatment of me has been extraordinarily fair despite my tremendous success? Do I just post it in a public forum?
As to further hints on how to model this, I would point out (and you probably already know this) that the problem with regression is the big standard error that is introdcued when you try to model using such a small sample size. For example, trying to model how a -180 favorite will fair after four innings in a tied game can produce screwy results unless you have many thousands of samples. Simple logic says that the answer should be much higher than 50% and at least a little bit less than the approximately 64% chance they had at the outset. That is, unless there is a serious imbalance between one team's starting pitching strength vs. their bullpen's strength. For example, if the Braves are starting a AAA pitcher vs. Milwaukee's Ace that price might actually rise in a tie game as the game progresses.
Perhaps an easier thing to model would be each team's potential for future scoring....  |