Some thoughts... >> I've seen quite a few statistical studies that address the >> topic "can previous history help us project future player statistics". That >> question is asked on an individual level (How many TDs will Priest >> Holmes score this year?) and a general level (How many more >> yards will the 5th best WR have than the 25th best?) I'm not convinced that >> statistics can help us much there - it seems that each year is >> extremely variable.
I don't know much about fantasy football, but regardless of how 'variable' each year is, this seems like something that Bayesians and machine learning folks would have fun with. >> However, I think statistical analysis has more to teach us about the >> best way to build a team, based on projected stats. FFL orginally was >> about the art of building the best team, but with certain assumptions >> put in place, I think it can be a science. >> Kind of like building the 'best model' ... with certain assumptions. >> And believe me, there are plenty of variables (injuries, >> free agency) that keep FFL very interesting year after year. (Just ask the >> millions of FFL owners that spend hundreds of hours crunching the >> numbers every year.) . >> Hmmm, sounds like prior information and hidden variables to me. It smells Bayesian, looks Bayesian, ...it's obviously a duck. cheeriOs, pradyumna . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
