tag:blogger.com,1999:blog-38600807.post1501733899385626193..comments2008-10-16T11:58:52.842-04:00Brian Burkenoreply@blogger.comBlogger9125tag:blogger.com,1999:blog-38600807.post-86972104919435975912008-10-16T11:58:00.000-04:002008-10-16T11:58:00.000-04:00SR-You might be right about that. That's one of the criticisms of logit regression in general. I've already added in pretty strong 'regression to the mean' corrections so that the game probabilities aren't overconfident based on short term outlier team performance.<BR/><BR/>One way to check for overconfidence is to calculate the theoretical predicted accuracy of the regression and compare it to the actual accuracy. I've only done that for one season, 2007, and it was dead-on--72% predicted and 72% actual. So I'm not sure if further corrections are wise.Brian Burkehttp://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-80039823544071831632008-10-16T03:55:00.000-04:002008-10-16T03:55:00.000-04:00Do you think that the win probabilities are accurate at the tails? I find it hard to believe that any NFL team has a 5% or less chance of beating any other. Yes, I thought this even before the Rams and Browns won.Sexy Rexynoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-62893684305690992822008-10-15T13:30:00.001-04:002008-10-15T13:30:00.001-04:00Looks like the Panthers moved up despite a poor showing simple because the Giants were ahead of them and had a poorer one.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-41229609836800459152008-10-15T13:30:00.000-04:002008-10-15T13:30:00.000-04:00Re: the Panthers. It could be a couple things--the teams they played previously may have done very well last week, improving their SoS. Also, even thought they lost, they might not have had terrible "internals" or things my model doesn't capture like punt returns or missed fld goals. Another thing could be that the team(s) ahead of them did even worse.<BR/><BR/>Re:Washington vs avg team on a neutral site. Don't forget the constant. It's about -.3 or something. That might fix it.Brian Burkehttp://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-41502284716507522412008-10-15T12:07:00.000-04:002008-10-15T12:07:00.000-04:00There is something I don't understand in the model. I ran the regression with last year coefficients, and set the AHOME at 0.5 for neutral site. Then I took Washington's stats for the "A" coefficients, and the average's stats for the "B" coefficients. I found a GWP of 77% without adjustments. Then I reversed the A and B stats and found a GWP of 21%.<BR/><BR/>So Washington playing an average team on neutral site has a GWP of 77% but an average team playing Washington on a neutral site has a GWP of 21%. Am I missing something?<BR/><BR/>Second remark on this, why both percentages don't give 100%.Sports Picks Systemhttp://www.sportspickssystem.comnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-60889088286508492432008-10-15T10:37:00.000-04:002008-10-15T10:37:00.000-04:00More surprising than the Skins staying on top was the Panthers moving up a notch after an awful showing in Tampa. What happened there?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-81642463801904833482008-10-15T02:17:00.000-04:002008-10-15T02:17:00.000-04:00I was surprised too. Thought they would fall dramatically, but they didn't for 3 main reasons: They still have pretty good eff stats, they still have zero ints and a very low fumble rate, and they have had one of the toughest schedules.Brian Burkehttp://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-17962371296951566982008-10-14T20:44:00.000-04:002008-10-14T20:44:00.000-04:00Why do you think the Rams stay at #1 despite losing to the 29th best team (from last week's rankings?) Was that loss not as bad as it seemed (not sure how) or is there something else going on here?Chasewww.footballguys.comnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-50469377777149474712008-10-14T20:39:00.000-04:002008-10-14T20:39:00.000-04:00Oops. Just click on the Rank header to sort in order of team strength.Brian Burkehttp://www.blogger.com/profile/12371470711365236987noreply@blogger.com