tag:blogger.com,1999:blog-38600807.post88358696420117240..comments2023-11-05T04:16:44.937-05:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Momentum Part 2: The Effect of Momentum-Swinging Events on Game OutcomesUnknownnoreply@blogger.comBlogger10125tag:blogger.com,1999:blog-38600807.post-73048810883015954742013-11-24T05:53:29.718-05:002013-11-24T05:53:29.718-05:00Possessions represent an opportunity to score. Don...Possessions represent an opportunity to score. Don't think of them as drives so much as "nonzero scoring probabilities," where advancing increases the probability to score.<br /><br />A turnover reduces that probability to zero for the opposing team and creates a nonzero, likely significant probability for you.<br /><br />I imagine that the relationship between turnovers/wins and punts/wins is in fact not that far apart (or as far apart as narratives would have you believe). If we compared teams by the number of punts they were forced to kick, I doubt we'd end up with a much different narrative (although turnovers do a better job in terms of field position, usually, then punts).KernelReeferhttps://www.blogger.com/profile/14284097281238034150noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-75234082184771076092013-11-22T20:00:00.979-05:002013-11-22T20:00:00.979-05:00Since possession alternates, I don't see how a...Since possession alternates, I don't see how an extra possession results. Certainly a shorter one but if the game impact is the same as punt plus return, why the correlation with wins?JMMnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-44217397311008063272013-11-22T12:46:07.526-05:002013-11-22T12:46:07.526-05:00Turnovers and winning are correlated because: (1)...Turnovers and winning are correlated because: (1) turnovers create more drives for your team, resulting in more points for you; (2) turnovers take drives away from the opponent, leading to fewer points for them; and (3) they potentially indicate a general difference in skill between teams.<br /><br />What Brian is saying with the momentum analysis is that, if a drive begins after a turnover, that drive is no more likely to lead to points than an average drive that begins in the same location.Brian Andersonnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-88084827683374330232013-11-22T08:31:44.766-05:002013-11-22T08:31:44.766-05:00First, I think this is an interesting set of artic...First, I think this is an interesting set of articles. Brian, you mentioned in your comments an upcoming article on streaks. I have been wondering how to use the change in slope from negative to positive, or vice-versa, as the foundation and definition of momentum. I am looking forward to the next installment. <br /><br />There is one element which I still have trouble reconciling, if how one gets the ball doesn't matter then only where and when one gets the ball matters. Yet there is a very high correlation between turnovers and winning. Do teams who are already behind turn the ball over more by taking more risk? Is the correlation not as strong as reported?<br /><br />I don't think this gets put behind us (me) until we (I) know.<br /><br />The last point is I have a nagging feeling that looking for momentum in game stats is somehow analogous to looking for thoughts in a brain during autopsy. If you use the autopsy results to draw your conclusion, you can conclude thoughts don't reside in the brain, yet we all think they do.JMMnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-29765093125170410352013-11-21T14:00:42.953-05:002013-11-21T14:00:42.953-05:00Nice post. Interesting that when only looking at 1...Nice post. Interesting that when only looking at 1st-3rd quarter situations the stop on downs cases appear to drop even farther below the non-momentus cases as compared to looking at all cases. This probably means that when looking at ONLY 4th quarter situations the stop on downs might actually result in higher win probabilities than the non-momentus cases (at least hinting at the possibility of momentum). Wouldnt a stop on downs (theoretically) be more of a momentum swinger in the 4th quarter than during the first three? And wouldnt a (theoretical) momentum swing late in the game have a bigger effect on game outcome than one much earlier where a variety of other factors could wash it out? Id be interested in seeing this graph just for late game scenarios when the win probability is closer to 50/50 (as mentioned in the other posts). Other course there is no evidence of this for any of the other momentum swinging events, so there's that... <br /><br />p-values!Torin Clarkhttps://www.blogger.com/profile/15151257492957093380noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-73307592080456728722013-11-21T13:18:33.479-05:002013-11-21T13:18:33.479-05:00Brian A. - I agree. But it's not a requirement...Brian A. - I agree. But it's not a requirement that all of those plays be momentum-swinging' for us to detect momentum. If just some of them are, we'd expect to see some sort of bump in win%.<br /><br />To eliminate the non-significant 'false' momentous plays, it would be very easy to include only those that take place when the WP is reasonably close to .50. That's when those kinds of plays are high-leverage. Or I can actually use my Leverage Index stat that's been gathering dust for yrs.<br /><br />Andy- Good points, especially regarding consecutive conversions on drives. I think one of the upcoming posts will address that head on by measuring how 'streaky' games are.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-28423452759633754542013-11-21T12:53:21.550-05:002013-11-21T12:53:21.550-05:00In human performance, as in physics, momentum is c...In human performance, as in physics, momentum is conserved until it encounters a force that reduces or negates it. In football, that force can be an external defensive counter, or an offensive strategic failure.BillBagleyhttps://www.blogger.com/profile/05013617543516721262noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-80828431874296717662013-11-21T12:48:19.917-05:002013-11-21T12:48:19.917-05:00One thing that may be working against this analysi...One thing that may be working against this analysis (and the first one) is that not all of the events you've labeled "momentus" are actually momentus. For example, suppose the defensive team is winning big, and they intercept a pass. It's hard to argue that play causes momentum for the team that is already winning big. I'd think you'd want to filter those out.<br /><br />One idea is to define "momentus" plays as those that cause a reasonably large change in winning probability, and then perform the same analysis as above. You'd want to see if large swings in win probability from a single event lead to different "true" win percentages than the average game.Brian Andersonnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-33437177047064878272013-11-21T11:33:47.415-05:002013-11-21T11:33:47.415-05:00This sounds an awful lot like the "Team Ident...This sounds an awful lot like the "Team Identity" post(s) Brian had last year. Some crap that announcers spew out because they can't give any real analysis. Team A just isn't playing well because Team B is more talented, better coached, or getting lucky. I'd rather have an announcer tell me that than make up momentum or identity crap. Apparently, that's too difficult for most.stevekirschhttps://www.blogger.com/profile/18140362048199788298noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-141517926049669812013-11-21T10:05:03.105-05:002013-11-21T10:05:03.105-05:00I think people have mentioned this earlier in comm...I think people have mentioned this earlier in comments, but it seems like the last place momentum can hide is when a team sustains a drive. I think this analysis clearly shows that whenever there is a change of possession there probably isn't any momentum to be carried over. But what a about when one team keeps the ball? <br />One commenter had the idea to compare subsequent 1st down conversion rates after a team has already obtained a previous 1st down on that drive. This could probably be done more rigorously with expected points. <br />There are some effects that would have to be weeded out; good teams are more likely to get a first down than the 66% leauge wide conversion rate - so when they go ahead and get that first down, obviously they will continue to convert first downs at their true rate, not the league wide average.<br /><br />Heres the kicker - even if an effect like this is seen (higher than average 1st down rates), it isn't necessarily "momentum" how we think of it - it could just be an offense exploiting a specific weakness. This would last as long as it takes the defense to recognize and react. <br /><br />Good example - Colts vs. Saints in the super bowl. Manning and Wayne ran that 5 yard square in like 10 times (almost) in a row. It was automatic, until Porter stepped in front and returned it 70 yards for a touchdown. I wouldn't call those previous completions momentum, it was just the colts exploiting an area of the field that the Saints weren't covering very well...until things changed.Andyhttps://www.blogger.com/profile/13033784825091918451noreply@blogger.com