tag:blogger.com,1999:blog-38600807.post8552238684144570770..comments2023-11-05T04:16:44.937-05:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Team Rankings: Week 6Unknownnoreply@blogger.comBlogger33125tag:blogger.com,1999:blog-38600807.post-49678619436881284912011-10-13T06:08:53.685-04:002011-10-13T06:08:53.685-04:00Anonymous, I believe Brian will have the Dallas ga...Anonymous, I believe Brian will have the Dallas game this week as more or less a toss up, given home field advantage to New England, and you're right that this is some distance from the market view, which is that New England has about a 70% chance of winning.<br />As to why this is the case, as far as I know efficiency stats are not used to any great deal by any other predictor. There is no right or wrong in this, Brian's efficiency stats correlate well, but so too do points and yards when interpreted in the right way.Tomnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-38366736889572170352011-10-12T22:55:48.501-04:002011-10-12T22:55:48.501-04:00Ian S, I think most readers of this site are smart...Ian S, I think most readers of this site are smart enough to know that the best team doesn't always win and that the team with the best record isn't always better in a predictive sense.<br /><br />But I think it's a fair question to ask why this statistical model is any better than others, e.g. FO, CHFF, Sagarin, Accuscore, numberFire, Massey, Dunkel, and dozens of other computer models. As far as I know, no independent source is tracking and publishing the results of Brian's model (although it may be the most open).<br /><br />The Dallas ranking is particularly interesting because they play New England this week and it appears that Brian's model is projecting the Cowboys as the most likely winner. As far as I can tell, it is the only published computer model that doesn't have New England as a huge favorite and in that sense it is a major outlier. That doesn't necessarily make Brian's model wrong, but it make you wonder what data one model weighs so differently than all the others and whether there's any new knowledge that can be gleaned from it.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-91181782459559009062011-10-12T16:55:49.601-04:002011-10-12T16:55:49.601-04:00jimaron, Brian's post on luck and determinism ...jimaron, Brian's post on luck and determinism can be very quickly and succinctly put: the favourite in a football game varies from having a 0.5 chance of winning to a 1.0 (certain) chance of winning, so assuming that the probabilities across a season are uniformly spread, there is a 0.75 chance of favourites winning on average. However, the truth is that the strength of teams in the NFL is not uniformly distributed, it is grouped slightly around the mean of 0.5 GWP, and as such favourites win more like 66% of games. I dare say the limit on picking winners is around 70% rather than 75%, looking at the spread of strengths.Tomnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-50769399785971884222011-10-12T16:12:22.757-04:002011-10-12T16:12:22.757-04:00jmaron-Not quite. Teams can have greater than a 0....jmaron-Not quite. Teams can have greater than a 0.75 GWP. The ~50% luck number is for the league as a whole. When two perfectly equally-matched teams play each other, the outcome is essentially 100% random. (Over infinitely many theoretical games, each team will win half. We never know which potential outcome we'll be treated to.) So for every game in which the teams are very close in strength, there is going to be games in which one team is a heavy favorite.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-52480991822787917812011-10-12T15:57:14.411-04:002011-10-12T15:57:14.411-04:00Hi Brian,
just read your article on pct of luck ...Hi Brian, <br /><br />just read your article on pct of luck in football - which basically asserts roughly 50% of game outcomes is pure luck. Can I draw from that that no team would ever have a higher rating that .75 GWP? Or am I missing something?jmaronnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-75831286191255236882011-10-12T13:02:11.586-04:002011-10-12T13:02:11.586-04:00Joseph,
Oh yes, Y/PP is the most important predic...Joseph,<br /><br />Oh yes, Y/PP is the most important predictive Stat, while TO´s are mostly random (= almost non predictive). Brian has written very good articles about that. Actually it´s enough to rank offenses and defenses solely on Y/PP to come up with good predictive Rankings.<br /><br />Karl, GermanyAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-45925577408469764872011-10-12T12:15:58.438-04:002011-10-12T12:15:58.438-04:00I'm not going to knock the formula in general,...I'm not going to knock the formula in general, but I will ask one thing based on the coeficients--if OINT% is -13.xxx, and DAL is 29th (or is it 30th) and OFUM% is -7.xxx, and DAL is 20th in that, then how is DAL's offense 2nd? Are they light years better than other offenses in other categories? (In passing, they are top-5--in run success rate, their 29th also.)<br />So Brian, what gives? If they are horrible in 2 of 4 offensive categories, and bad in a third, how are they the 2nd overall offense? My Saints are better in the three categories where DAL is bad, and not much worse in OPASS category--yet they're the 5th rated offense. (I don't dispute that they're the 5th best offense--I just have a hard time believing that .5 yds per pass overcomes NO's large difference in both turnover categories.)Josephnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-41826662454268037182011-10-12T11:55:21.666-04:002011-10-12T11:55:21.666-04:00"DAL 1st? A 2-2 ahead of two 5-0 teams, ONE....."DAL 1st? A 2-2 ahead of two 5-0 teams, ONE..."<br /><br />If you really did read this site for two yrs and come up with such a post, it´s a shame for your intelligence.<br /><br />We had all this 1-time posting guys (trolls) here last year clobbering Brian for ranking ATL low. And what happened? They got (rightfully and predictable) destroyed in the playoffs. Brian´s model or even simple Y/PP showed that ATL were major overachivers (GB was exactly the other way around; and it really shows this season). The model identfies such teams. For now DAL is underachieving. If you don´t get it and/or like it, go read Prisco´s "Power-Rankings" and troll around there...<br /><br />Karl, GermanyAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-71421452934515814222011-10-12T11:19:14.120-04:002011-10-12T11:19:14.120-04:00Think I may have to make a copy and paste response...Think I may have to make a copy and paste response for every time someone comes on saying “oh these rankings are rubbish because you’ve got a 2-2 team above 5-0 teams”.<br /><br />If you want to see teams ranked by their record, go look at the standings. If you want to see teams ranked by how they are performing against the statistics that correlate with future wins, then these are a pretty good set of numbers for that purpose. Also please note that the NFL is not transitive i.e. if team A beats team B, and team B beats team C that doesn’t mean team A will beat team C. As for pointing out that DAL lost to DET, so DET must be better than DAL, remember the 16-0 Patriots? They beat the Giants in week 17, yet who did they lose to in the Superbowl?<br /><br />If you’re going to come on and rubbish these ratings then that’s fine, but use proper arguments rather than basing your opinion on a sample size of 1.Ian Snoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-76775559891078885582011-10-12T10:23:25.113-04:002011-10-12T10:23:25.113-04:00Anonymous, who is it exactly that you are getting ...Anonymous, who is it exactly that you are getting cross at here? It isn't like these represent a person's opinion and you happen to disagree with them, it in fact represents the outcome of a body of work that was performed in order to discover which statistics are most predictive of a team's chances of winning. It is not like Brian sits there each week and orders the teams, they have ordered themselves. Hell, you can hardly even say Brian chose the criteria, he simply tested the criteria against historical data to find which pieces were most predictive. He chose repeatable common instance statistics so that even though there are few games in a season, the information he uses is over far more data points than the number of games (or points), helping to remove an amount of noise in the data.<br /><br />No model is perfect, but Brian's has a key factor that your simple thought processes do not, it has convincing evidence.Tomnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-5713981273427151652011-10-12T09:05:26.806-04:002011-10-12T09:05:26.806-04:00Anon, I didn't realize reading the defense and...Anon, I didn't realize reading the defense and countering with an audible to the perfect play before firing a pass downfield was so "emotionally" complex. Maybe the reason the Bears offensive line can't block for Cutler is because they feel sad.<br /><br />Fun Fact: The score is not the only RESULT of a game. The Cowboys passed for 6.8 ypa vs the Lions 5.6 (56 vs 42% SR) and rushed for 4.2 ypc to the Lions 3.7, statistics that indicate the relative strength of the team's offense and defense, all of which favor the Cowboys.<br /><br />The Cowboys were also -2 in turnovers in a 4 point game. If you are an avid reader of this site like you claim, then you know that turnovers explain past results very well, but are not predictive of future results. Let's replay that game again 100 times and see how many interceptions, much less multiple pick-6 games, Romo has.<br /><br />Furthermore, let's talk about undefeated teams. The 49ers and Redskins have two combined losses this year, both of which are to the Cowboys. Had the Cowboys lost those games, then three of the four teams the Cowboys have played would be undefeated this year, with a combined 14-0 record. Instead, the Cowboys beat two of three undefeated teams. If you won't accept that as a signal of strength (which Brian's Opp GWP shows with much higher accuracy) then I don't know what else to tell you.Jameshttps://www.blogger.com/profile/01838293735141324662noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-69741849631968608202011-10-12T07:35:19.008-04:002011-10-12T07:35:19.008-04:00I am a HUGE Stats nerd and read your site for two ...I am a HUGE Stats nerd and read your site for two years now as this is fun, yet this is why I hate trying to "rank" something as emotionally complex as football which based on MANY other factors besides stuff a few basic stuff like runSR%. <br /><br />DAL 1st? A 2-2 ahead of two 5-0 teams, ONE of which is a team that actually BEAT Dal! Yeah, yeah I know, this is where I get the speech about the score is only a snapshot of a chosen ending to an arbitrary amount of time. But reality check--We have a RESULT and yet somehow some super formulas compute some chosen factors to ignore that factual result and come to the conclusion Dal is better by a pretty good margin (#1 vs #8).<br /><br />And JAX is the 2nd best DEF? LOL. Having prevent the other team from scoring in football would be common sense for a def, right? Well JAX has not allowed an opponent to score LESS than 14 points yet. And it allowed over 30 points twice in 5 weeks, one of which to the pathetic offense of the NYJ's, your 23rd ranked offense!<br /><br />Keep in mind also that this rating system is the same one that had GB BARELY in top 15 at this same point last year.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-68106050276047932622011-10-12T01:54:41.819-04:002011-10-12T01:54:41.819-04:00I assume it's safe to say Dallas is a victim o...I assume it's safe to say Dallas is a victim of variance and will likely make the postseason comfortably?<br /><br />Tampa bay might piece together some luck, the giants as well. Philly might play better, but right now I think it's safe to say Detroit, San francisco, new orleans, Washington, green bay, and Dallas have been the six best teams in the NFC.Joe r.noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-51608570250573608552011-10-12T00:50:43.452-04:002011-10-12T00:50:43.452-04:00Anon, I understand how you feel. It is difficult t...Anon, I understand how you feel. It is difficult to believe information that contradicts public opinion. Especially when hit with information from other sites and opinions from "experts" in the media. However, the result that Brian has achieved with his model in past seasons has been very solid.<br /><br />Note on ESPN Power Rankings: It looks like these rankings are determined by 5 people who vote on the rankings. One problem with this method is that the voters could feel compelled to vote with in-line with their peers or public opinion in order to avoid looking foolish.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-79083666891442154212011-10-11T22:22:36.610-04:002011-10-11T22:22:36.610-04:00Anon- At least take time to read how the rankings ...Anon- At least take time to read how the rankings work before asking that question. Football outsiders weights a lot of non-predictive events in DVOA.<br /><br />Take away just 2 Romo plays and the Cowboys are likely 4-0 and not 2-2. <br /><br />As has been explained numerous times win-loss record isn't always an accurate reflection of the innate team quality. And luck plays a big part in the outcome of games.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-75181576948631629662011-10-11T22:13:25.002-04:002011-10-11T22:13:25.002-04:00This might sound cranky. If so, I apologize. You a...This might sound cranky. If so, I apologize. You ask a very fair question. It's just tiring answering the similar questions every week when the answers are already provided.<br /><br />ESPN's power rankings? Seriously?<br /><br />As for DVOA, it's a measure of how well teams have played in the past. It's a black box based on tortured formulas designed to correspond with what everyone thinks already anyway. It's constantly chasing the previous year's numbers. (That should get the trolls going!)<br /><br />Do I personally think Dallas is the best team in the NFL? That's irrelevant. <br /><br />Do I believe that Dallas is the best team in terms of the facets of the game that are most predictive of future performance? So far, yes.<br /><br />Is that likely to change? Yes, to some degree.<br /><br />Should everyone take a second look at their opinion of Dallas? Absolutely.<br /><br />Would Dallas be 4-0 except for a couple very poorly timed turnovers? Yes.<br /><br />Besides, I shouldn't even have to answer these kinds of questions when all the necessary data is right there in the tables above.<br /><br />Off passing efficiency: 3rd<br />Def passing efficiency: 7th<br />Def run SR: 1st<br />To-date strength of schedule: 2nd<br /><br />Again, sorry for the crankiness. Fair question.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-49053118678537990162011-10-11T21:05:24.733-04:002011-10-11T21:05:24.733-04:00Football Outsiders has the Cowboys as the 17th bes...Football Outsiders has the Cowboys as the 17th best team in football right now. ESPN's power rankings have the Cowboys at 15. Your site, however, has the Cowboys as the best team in the league. Obviously, there is a huge discrepancy between the sites. Do you honestly believe that the Cowboys are that good and that everyone is just missing it? What do you expect to see from the Cowboys from this point on?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-29421286077060100712011-10-11T19:48:22.018-04:002011-10-11T19:48:22.018-04:00Ah, I assumed you were still using separate coeffi...Ah, I assumed you were still using separate coefficients as you did originally.Tomnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-45077757165521900222011-10-11T19:47:24.165-04:002011-10-11T19:47:24.165-04:00Mike B -- Regarding video tracking of individual p...Mike B -- Regarding video tracking of individual players, the product exists and several NFL teams are interested. "The league, as well as several individual teams (especially the Cowboys, Eagles, Falcons, Saints and Seahawks), are very interested in a motion capture technology called SportVU that can re-shape the way football is interpreted. The system, which is owned by Stats LLC, has created a quiet buzz in NFL circles."<br /><br />http://www.nationalfootballpost.com/NFP-Sunday-Blitz-1909.htmlAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-82019520412551286032011-10-11T19:44:54.757-04:002011-10-11T19:44:54.757-04:00Those are for both teams, except the visitor is in...Those are for both teams, except the visitor is inverse of the coefficients listed. There is a dummy for the home team: 1 if home and 0 if away. The model's result will be the probability of a home win.<br /><br />In other words, the coefficients are for computing each team's logit value. Subtract the visitor logit from the home logit, then add the home field dummy, then put it through the logistic wash.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-59142271497552660002011-10-11T19:23:10.343-04:002011-10-11T19:23:10.343-04:00Brian, are those not only the coefficients for the...Brian, are those not only the coefficients for the home team?Tomnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-62444579092324875372011-10-11T17:18:49.029-04:002011-10-11T17:18:49.029-04:00Take those probabilities with a grain of salt. The...Take those probabilities with a grain of salt. They all assume teams keep playing at a steady (regressed) level. But it doesn't regress for catastrophic injury--say a Calvin Johnson injury. <br /><br />It does look pretty clear the NFC is a conference of haves and have nots.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-38733635246192756492011-10-11T16:43:59.300-04:002011-10-11T16:43:59.300-04:00Looking at the software provided at nfl-forecast.c...Looking at the software provided at nfl-forecast.com (which is based upon Brian's GWP), it projects the NFC playoff teams are nearly set already. 3 teams are above 95% for the playoffs!<br /><br />Packers - 96%<br />Saints - 98%<br />Lions - 95%<br />Cowboys - 83%<br />49ers - 89%<br />Redskins - 60%<br />Giants - 30%<br /><br />While there's only a 40% chance that the top 6 teams will represent the NFC, it's an extreme contrast to the AFC. Not a single AFC team has playoff probabilities above 80%, and there's only a 12% chance it's current top 6 teams make the playoffs.<br /><br />Without thinking about it more I'm not sure what that means yet (AFC is deeper? NFC teams are more dominant?), but I thought it was worth sharing.Jameshttps://www.blogger.com/profile/01838293735141324662noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-17725666670645749912011-10-11T15:42:00.398-04:002011-10-11T15:42:00.398-04:00Just wanted to let you know I just discovered your...Just wanted to let you know I just discovered your site last week, and I absolutely love it. Keep up the good work!<br /><br />Especially enjoyed the "What Makes Teams Win" and "Luck and NFL Outcomes"Dustin Bnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-14059817078093441242011-10-11T15:33:42.479-04:002011-10-11T15:33:42.479-04:00$C['const'] = -0.4619; ...$C['const'] = -0.4619; <br /> $C['ahome'] =0.748577; <br /> $C['aopass'] = 0.4467;<br /> $C['aorunSR'] = 0.03565;<br /> $C['adpass'] = -0.576;<br /> $C['adrunSR'] = 0.02749;<br /> $C['aoint'] = -13.026;;<br /> $C['adint'] = 14.24;<br /> $C['aofum'] = -7.209;<br /> $C['apen'] = -1.587;Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.com