Converting Baseball predicted team totals to lines

[gamble4heisman]

Have their been any threads on this. Recently made a model predicting team totals on a per game basis but am struggling on converting them to my own predicted lines.

Any ideas or points in the right direction are much appreciated.

thanks,

G4H

[thom321]

Here is my crack at it, for calculating money lines. Assuming your model project Team A to score 5 runs and Team B to score 3 runs, you can use the Pytagorean formula to calculate win %. Using a coefficient of 1.83, the calculated win probability for team A would be (5^1.83)/(5^1.83 + 3^1.83) = 71.805%. The win probability for Team B would hence be 1 - 71.805% = 28.195%. Check out the wiki for more details and info on adjustments to the formula.

http://en.wikipedia.org/wiki/Pythagorean_expectation

[mathdotcom]

Your model should give you run distributions for each team from which you can calculate Pr(A runs > B runs) (giving you a fair ML on team A)

[buby74]

Try the tango distribution for coverting a team average to a run distribution

[gamble4heisman]

thom, does the 1.83 factor hold for creating lines for first five innings?

[thom321]

gamble4,

I don't know and I don't want to assume anything either. The 1.83 coeff was obviously not something I came up with myself. However, there is quite a bit written about the formula on the Web so I would google around a bit and see what others have said.

[TomG]

The exponent depends upon the scoring environment. Adjustments are needed.

[TomG]

http://www.baseballprospectus.com/article.php?articleid=342

[evo34]

Also worth reading:

http://web.williams.edu/go/math/sjmiller/public_html/math/papers/PythagWonLoss_Paper.pdf

[uva3021]

use the run scoring dependent exponent

(rs+ra)^.287