tag:blogger.com,1999:blog-38600807.post3105297820322860475..comments2018-06-02T14:19:34.554-04:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Team Playoff Probabilities - Week 11Unknownnoreply@blogger.comBlogger13125tag:blogger.com,1999:blog-38600807.post-53964766902318785702009-11-25T01:21:29.496-05:002009-11-25T01:21:29.496-05:00zlionsfan,
Thank you for that clarification--I def...zlionsfan,<br />Thank you for that clarification--I definitely get the madden comparision and why this sort of thing would never predict an Oakland win. I guess its the games like Arizona, Phili, Denver or even Tenn (where the teams are different game to game or from bad to better, or vice versa). Anyway, the thanks for that explanation--I will try to apply these things when making my pics and think I have a better idea of how to now.Heathernoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-62112753930413514642009-11-24T18:28:15.025-05:002009-11-24T18:28:15.025-05:00I would say a couple of factors: teams whose perfo...I would say a couple of factors: teams whose performances have risen and fallen significantly during the season (like Arizona and Philadelphia) and teams whose performance has essentially taken a U-turn at some point in the season (like Denver and Tennessee). <br /><br />Teams in the former category are difficult to predict because a system probably can't determine "which team will show up" on a given week. Teams in the latter category are difficult to predict a) when the U-turn is made, if it's caused by specific events (injuries, coaching changes), and b) for a while after the U-turn is made, because a system probably can't determine whether a given team is a crazy team or a U-turn team. <br /><br />As far as performance in an individual week, some of it is caused by the types of teams above, and some is caused by aberrant performances. Looking at relative strengths of teams, a system can't really say it thinks you should take Oakland over Cincinnati; if the system suggested Detroit over Green Bay on Thursday, you'd be right to question it. If Oakland does beat Cincinnati (or Detroit over Green Bay), sometimes it's simply an upset. <br /><br />After all, given the percentages above, the system did give U-turn Tennessee a 3 in 10 chance of beating Houston. Even Oakland's 1 in 8 shot still means, well, 1 time in 8, and this week happened to be that 1. <br /><br />I think systems would be more easily understood if they were applied to, say, Madden. Don't laugh, here's what I mean: in that environment, you could "play" the games 100 times, 1000 times, or whatever to overcome the single-event problems we have in evaluating systems vs. real life. There's no way to replay the Cincinnati-Oakland game 99 times under the same circumstances to show that look, the Bengals won 84 of them, that's pretty close to 0.87-0.13.zlionsfanhttps://www.blogger.com/profile/02966540737106797756noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-4238993741917195932009-11-24T01:06:27.466-05:002009-11-24T01:06:27.466-05:00I have been following this blog for quite some tim...I have been following this blog for quite some time and was thinking that the weekly predictions were pretty good. But this week was a total wreck (Tennesse SO big over Houston as only one example). And how is it that Arizona keeps having such a small probability of winning when they clearly are going to win? I know a lot about regressions and understand fully how these are done, but it seems there is something off....thoughts?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-87942155736332314442009-11-22T14:10:23.451-05:002009-11-22T14:10:23.451-05:00James-Thanks. That happens when there is an error ...James-Thanks. That happens when there is an error in the source data. I have a routine that is supposed to catch that, but it's not foolproof.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-59739362735412391602009-11-22T13:45:41.485-05:002009-11-22T13:45:41.485-05:00Hey, uh, the win probability chart randomly jumps ...Hey, uh, the win probability chart randomly jumps from halftime back to the start of the 2nd quarter in the Cowboys-Redskins game.Jameshttps://www.blogger.com/profile/01838293735141324662noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-63804204223525993812009-11-22T09:44:31.688-05:002009-11-22T09:44:31.688-05:00Here are the probabilities for the entire slate of...Here are the probabilities for the entire slate of Week 11 games. <br /><br />Win Chance GAME Win Chance <br />0.35 Miami at Carolina 0.65 <br />0.20 Washington at Dallas 0.80 <br />0.33 Cleveland at Detroit 0.67 <br />0.23 San Francisco at Green Bay 0.77 <br />0.90 Pittsburgh at Kansas City 0.10 <br />0.24 Atlanta at Giants 0.76 <br />0.91 New Orleans at Tampa Bay 0.09 <br />0.26 Buffalo at Jacksonville 0.74 <br />0.66 Indianapolis at Baltimore 0.34 <br />0.20 Seattle at Minnesota 0.80 <br />0.62 Arizona at St. Louis 0.38 <br />0.24 Jets at New England 0.76 <br />0.87 Cincinnati at Oakland 0.13 <br />0.41 San Diego at Denver 0.59 <br />0.68 Philadelphia at Chicago 0.32 <br />0.30 Tennessee at Houston 0.70 <br /><br /><br />Brian explained how his predictions work in this post.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-44352093448305557172009-11-21T17:18:57.643-05:002009-11-21T17:18:57.643-05:00Brian -- thanks for posting my work!
A few assump...Brian -- thanks for posting my work!<br /><br />A few assumptions that go into the season-long predictions:<br /><br />1. Each game is independent (e.g. catastrophic injuries not accounted for). As Brian mentioned.<br />2. The game results (statistics used by Brian) are a true representation of the team strength. In reality, at best the game results are a sample of the true team strength. This tends to become less significant by this stage of the season when many games have been played. But I start these predictions early in the season when the sample size is small and the game results may not be a true indication of team strength.<br />3. Team strength does not change significantly during the season. <br /><br />Assumptions 2 and 3 are kind of related to one another and difficult to unravel from each other during a relatively short NFL season. An example of 2 and/or 3 working together is that I had Denver as a 99.6% chance of winning the AFC West division earlier in the season (around week 6). This estimate looks to be too high, in hind sight. I may try to account for these weaknesses in future years.Chrishttp://nfl-forecast.comnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-53223093684515682292009-11-21T13:11:48.159-05:002009-11-21T13:11:48.159-05:00Independent--so something like a catastrophic inju...Independent--so something like a catastrophic injury is not factored in.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-49687349524914160242009-11-21T12:05:26.786-05:002009-11-21T12:05:26.786-05:00Can I ask if the model treats every game as an ind...Can I ask if the model treats every game as an independent event or does it account for the fact that losing one game could affect the probabilities for the next one? For example, Indianapolis has a 34% chance of losing this week. They'll probably have a 30% chance of losing the next week. Is the probability of losing both weeks 0.34 * 0.3? I'm not sure if this accounts for the fact that the same event could cause them to lose both weeks. For example, what happens if Manning got injured? I think this possibility is accounted for in both individual probabilities, but the cumulative effect isn't considered. Am I wrong on this?Jeff Clarkenoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-84924081777246299862009-11-21T11:42:00.687-05:002009-11-21T11:42:00.687-05:00It's possible that if these were rounded to te...It's possible that if these were rounded to tenths or hundreths of a percent, some of those teams would still have a chance. But it's not that surprising that they made the playoffs less than 1 time in 200 in simulation.Tarrhttps://www.blogger.com/profile/14368810359650066790noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-59221231801606182212009-11-21T10:56:01.532-05:002009-11-21T10:56:01.532-05:00Shaun,
Detroit, Cleveland and Tampa Bay are all t...Shaun,<br /><br />Detroit, Cleveland and Tampa Bay are all two losses away from elimination. <br /><br />Many of the teams on the bottom are three or four games away from elimination (note this includes a wins by a division leader, and losses by the teams themselves).<br /><br />Since this model takes into account team efficiency ratings and most the teams on the bottom are rated poorly it makes sense for there to be a lot of 0's.<br /><br />Here is a link to a site that does not take into account team efficiencies. The game play is purely 50-50 so while it is not as accurate as the simulation shown above it may be nice to look at if you are rooting for a losing team.<br /><br />http://www.sportsclubstats.com/NFL2.html<br /><br />Hope that helps!mustardboy3141https://www.blogger.com/profile/12147464683064864133noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-45825370353096303142009-11-21T10:54:34.154-05:002009-11-21T10:54:34.154-05:00This comment has been removed by the author.mustardboy3141https://www.blogger.com/profile/12147464683064864133noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-67461041610652067752009-11-21T06:12:08.582-05:002009-11-21T06:12:08.582-05:00Thats a lot of 0's considering how early we ar...Thats a lot of 0's considering how early we are in the season. Surely these teams cant be written off completely so early.Shaunnoreply@blogger.com