Win percentage question

[runnershane14]

Lets see if you guys have any ideas for this. For example: If team A is expected to win 65% of all of their games this season and team B is expected to win 40% of all of their games on the season, is there a way to predict how likely team A is to beat team B in a game. Any thoughts?

[Peregrine Stoop]

no... unless they had the same opponents left

[Justin7]

If 50% is the average team, you want an equation where a 65% team playing a 50% team will win 65% of the time. It should go up against weaker teams, and down against better teams.

One "non-perfect" way is to give each team a ranking relative to 50 (65 is +15; 40 is -10). A 65 vs a 40 would win 50 + 15 - (-10), or 75%.

A different approach is to take some sort of geometric mean, forcing it into the above constraints.

[Thremp]

I stumbled through google in a drink induced haze until I just IMed a friend for the name of what this technique was. The things I do to help others: http://www.diamond-mind.com/articles/playoff2002.htm

http://www.tangotiger.net/wiki/index.php?title=Log5

[durito]

before he was fired ganch goes over the log5 method and others here:

http://forum.sbrforum.com/handicapper-think-tank/118483-setting-fair-line.html

[20Four7]

before he was fired ganch goes over the log5 method and others here:

http://forum.sbrforum.com/handicappe...fair-line.html

Ganch wasn't fired, he's on special assignment.......