|Clayton Kershaw - Vegas' Favorite Pitcher|
As with my other sports rankings, my goal is to reverse engineer an implied power ranking from the betting lines and totals. In other words, I'm trying to figure out what the betting market "thinks" are the best and worst teams.
Because baseball uses moneylines instead of run spreads, I have to convert the moneylines into an implied probability, and then combine that probability with the betting run total to arrive at an implied (fractional) runs scored for each teams (using pythagorean expectation). I can then run my regression analysis just as I do for the other point spread-based sports.
Another complication specific to baseball is the starting rotation. The Dodgers have a much different win probability with Clayton Kershaw on the mound than they do with Stephen Fife. So, my rankings have to treat each team as if it were five (and sometimes six) separate teams, one for each starter in the rotation. While that complicates things somewhat, it does allow me to derive an implied pitching ranking from the betting information.
Team rankings can be found here and are updated daily. As you can see from the sparkline, the Tigers have been at the top of the rankings all season. It's always interesting to see where the rankings diverge from the win-loss record. The Pirates, despite a #3 ranked win-loss record, are in the middle of the pack (#14) according to the betting lines.
These are also updated daily and can be found here. These are a bit more noisy than I would like, and I am exploring ways to improve accuracy and stability. Clayton Kershaw is currently at the top of the rankings. With Kershaw starting, a league average team would be expected to score 1.11 fewer runs than when facing a league average starter.
As with other sports, I will publish my Today's Games feature on a daily basis. The general idea is to use my rankings to predict the betting information for the day's games, and highlight where the market differs. Big discrepancies are often indicators of late breaking information, like injuries. However, misses are also a function of model error too, and I'm thinking that my model error noise may be drowning out any useful signal. Use at your own risk.
The rankings are on a ballpark neutral basis. See here for more background.