Page 1 of 1

Basing Ratings on Statistics

Posted: Thu Feb 01, 2007 9:20 am
by dp68
This has probably been discussed ad finitum but I haven't found an appropriate thread using Search.

Is there a good primer for creating a player's ratings based around a set of statistics? For instance how does a batter's batting average, k's vs bb's, and power numbers project into CH, PH, and Eye.

Thanks if you can point me in the right direction! [8D]

RE: Basing Ratings on Statistics

Posted: Fri Feb 02, 2007 1:20 am
by KG Erwin
dp68, I'm trying to get an explanation of how this works, as my belief is that ratings are relative to the entire database of available players for a given year. In other words, I THINK there's a "grading curve", and using absolute stats won't work. THIS is what I hope means grading players from different historical eras into one all-inclusive database, but somehow I don't think that "normalized" stats are figured in.

RE: Basing Ratings on Statistics

Posted: Fri Feb 02, 2007 5:25 am
by JonnieR
As I understand it the separate ratings are based on a player's stats relative to performance of all of the players. For instance if the "contact" rating is a function of the number of (At Bats - Strike outs)/At Bats, then you must calculate the mean and standard deviation for all the players in a league. Assuming a normal distribution you can then calculate an individual player's percentile ranking. Based upon the range for this particular rating, say 25 -100 for example you can calculate a player's ranking for making "contact". For instance a player with percentile of 75 (he is better at this rating than 74% of the players in the league), would have a contact rating of 81.

Theoretically every time you add a player to the database a rating may change because statistics for group change also.