Saturday, February 18, 2012

I already have the stats I want to use for the Pittsburgh Steelers and Baltimore Ravens. Win Percentage, Pass Yards,Run Yards, Turnovers, Etc.

Can I even use multiple regression to pick a winner or do I need more than two teams? I know how I can use multiple regression to check all thirty some teams to see who has the best chance to win a super bowl but what about just two teams.|||Just pick Baltimore.
The Steelers are the worst Super Bowl defenders ever.
Tongue in cheek...

My initial guess would be no, you cant do this, since each team does not play every other team, and even if they did, they wouldn't play them the same number of times, and even if they did that too, the sample space would still be too small-you would need dozens of games between any two teams.
The best analysis would be within the same division, but again, since they dont play the same teams out of division you'd have to use only the data in intradivision games and apply the results similarly.
I would also assume that the regression variables would be wildly non-linear, so you're going to have a beast of an equation when you're done.
Again, these are all just guesses by inspection, and they may be way off, but then again if it were in any way feasible I assume it would have already been done.
Nice try.|||Boy Steelers got crushed.
I dont know that any equation could have predicted that.

Report Abuse

|||I assume that you mean "multiple linear regression", which is not the ideal tool for this sort of problem. If you have statistical software available, then you might try logistic regression, which can be used to predict probabilities, such as the probability that the Steelers win.

If you don't have statistical software, you still might try this using linear regression, but I suggest using the point spread as the dependent variable.

Good luck!

火车采集器

No comments:

Post a Comment