tag:blogger.com,1999:blog-38600807.post7782468759869409821..comments2023-11-05T04:16:44.937-05:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Luckiest and Unluckiest Teams - Week 14Unknownnoreply@blogger.comBlogger23125tag:blogger.com,1999:blog-38600807.post-68334790222417564752009-12-15T13:04:27.630-05:002009-12-15T13:04:27.630-05:00Why is the difference between your model and reali...Why is the difference between your model and reality termed "luck?"<br /><br />Is it not just a reflection of the inaccuracy of the model?<br /><br />If you model was "perfect", then there would be no lucky teams. There would be no difference between actual and predicted performance. "Luck" by your definition would cease to exist if you had a better model.<br /><br />But the concept of "luck" or "luckiness" has nothing to do with your model or how well it models reality.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-90558597244145475172009-12-11T18:20:20.537-05:002009-12-11T18:20:20.537-05:002 things I don't like about the reasoning expr...2 things I don't like about the reasoning expressed in the original post:<br />1) Kick/punt returns for TD's aren't anymore lucky as to when they occur than any other long TD's; therefore, the Jets' missing "expected win" can be totally accounted for by their horrible kickoff return coverage in the Miami game; and<br />2)Made/missed kicks at crucial times are the key metric in evaluating the quality of a kicker...two kickers with identical made %'s overall but very different %'s with the game on the line are by definition good or bad kickers...a 16 game season is enough to judge that...<br />Jon SilverbergAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-50501220832496120322009-12-10T19:42:26.270-05:002009-12-10T19:42:26.270-05:00It's hard to deny the luck of the Saints and C...It's hard to deny the luck of the Saints and Colts this season. If you don't believe the stats, just look at the anecdotal evidence. How many come-from-behind victories does a team need before you call it luck? If this list is flawed, it is not because the best teams are at the top. That's just the way this season has been.<br /><br />However, there are certain players who make clutch plays week after week, year after year, like Manning and Favre, and off the top of my head I would add Troy Polomalu and the Freeney/Mathis duo on the defensive side.<br /><br />Notice, for example, how many Steelers losses this year have come on last-second defensive collapses in Polamalu's absence, and their win over the Titans was in no small part due to Polamalu's big plays in the clutch. I am wondering, Brian, if it would be possible to come up with some sort of individual clutch rating based on your existing data?Brettnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-81539595092659066402009-12-10T17:54:54.416-05:002009-12-10T17:54:54.416-05:00I don't have much time, so I'll try to ans...I don't have much time, so I'll try to answer as many questions as I can quickly.<br /><br />"Clutch" is definitely part of it. Whether someone can chose to be better than he otherwise is in the clutch is another discussion.<br /><br />This is based on GWP, which is adjusted for schedule strength. As I wrote in the original post, teams do not control their schedules. If a team wins a lot of games because they've been served up easy opponents, that's part of luck.<br /><br />The "expected" part of expected wins refers to the statistical expectation at this point in the season (12 games in). It doesn't mean I'd expect them to win that many by the end of the year.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-72925433634576598932009-12-10T16:57:02.348-05:002009-12-10T16:57:02.348-05:00So when you say a team is "lucky", is th...So when you say a team is "lucky", is this a good predictor of future events? Are the Vikings going to regress toward their GWP, or is it possible that they are in control of their luck? Which specific scenarios during a game count as lucky?Jackhttps://www.blogger.com/profile/03609393271186542173noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-91440694636563436562009-12-10T12:36:13.291-05:002009-12-10T12:36:13.291-05:00Not sure if this has been mentioned before, but is...Not sure if this has been mentioned before, but isn't it a problem that no team is expected (according to your model) to win 10 games? Does this indicate that all teams will likely win a few games due to luck, or are there simply typical game elements not being measured here? That said, I can certainly agree that the Vikings owe some of their success to luck (49ers game and Ravens game come immediately to mind).Unknownhttps://www.blogger.com/profile/10496187663354569243noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-71926510781348198452009-12-10T12:29:38.721-05:002009-12-10T12:29:38.721-05:00I see what anonymous is saying, even though he'...I see what anonymous is saying, even though he's confused. The top 4 teams are likely going to be the top 4 seeds in the playoffs. Additionally, the top 2 teams in terms of expected wins (IND and NO) are both in the top 4 for luck. However it stops there, after that, there is actually a slightly negative correlation between expected wins and luck.<br /><br />Anonymous should also remember that part of the reason we consider those top 4 teams to be excellent teams is their superior record, which is a result, in part, of their luck.Dan Schlauchhttps://www.blogger.com/profile/16890251806563865810noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-59162840592981592192009-12-10T12:15:08.925-05:002009-12-10T12:15:08.925-05:00Anonymous,
What makes you think the distribution ...Anonymous,<br /><br />What makes you think the distribution is not random? What clear pattern do you see?Natenoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-41384040376017472122009-12-10T11:53:47.366-05:002009-12-10T11:53:47.366-05:00If the residual were truly random "luck"...If the residual were truly random "luck" then the distribution of the error terms should be random. There shouldn't see any pattern in which teams are "luckiest" and which are "unluckiest". Given the rather clear pattern that we see among which teams are luckiest and which are unluckiest, I think you are either missing some relevant factors in your regression equation and/or have misfit the data (e.g. using a linear regression where a curvilinear one is needed). It is patentlly clear that the numbers you report are NOT simply luck.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-37815267452056684032009-12-10T11:31:57.667-05:002009-12-10T11:31:57.667-05:00I was curious about that too, Nate. I checked the...I was curious about that too, Nate. I checked the pearson correlation of of GWP and luck and found none (-0.001).<br />This at least suggests that any real elements missing from the model (special teams, clutchness, etc.) are not correlated with the quality of the teams, as I would have expected if something important was missing from the model.Dan Schlauchhttps://www.blogger.com/profile/16890251806563865810noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-31295609089470969092009-12-10T11:11:04.048-05:002009-12-10T11:11:04.048-05:00I'm curious about the answer to Zach's que...I'm curious about the answer to Zach's question too. Does expected wins factor in a team's GWP AND their opponent's GWP?Jameshttps://www.blogger.com/profile/01838293735141324662noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-84117673654509170472009-12-10T00:57:15.894-05:002009-12-10T00:57:15.894-05:00"Teams and players can't control when tho..."Teams and players can't control when those events occur, or else they'd save up their successes for when they matter most and their failures for when they don't."<br /><br />Precisely. Which is why calling the leader in WPA the MVP is a farce. Just because his timing happened to be best for that team doesn't mean he would have been as valuable for any other team in the league.<br /><br />Take for example a kicker that makes his first twenty kicks of the season and misses his next 5. It's likely his WPA grades lower than a kicker who missed five meaningless field goals, instead making the 20 most important. That's a function of luck; nothing more, nothing less. I'm sure there are actually some teams who, in that season, would have won more games if their team hit field goals 1-20 than a random spread of 20 over 25. It doesn't mean the first kicker is any less valuable than the other...the only information you know about those kickers that has any meaning is they're both 80% kickers in that season. That makes them equally valuable.<br /><br />Giving Favre the MVP award based on WPA and then turning around and basing a 'luck' article on team efficiency makes no sense. Either say Favre's the Most Lucky Player, or give the award to the player most influencing the most correlated statistic to consistent winning, ie. the quarterback of the team with the highest passing efficiency of the league.Marverhttp://pigskintelligence.blogspot.comnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-74253884468728385892009-12-09T22:33:49.288-05:002009-12-09T22:33:49.288-05:00I wonder if there's any correlation between a ...I wonder if there's any correlation between a team's luck during the first half of the season and their luck during the second half, or between one year and the next.<br /><br />If there's no correlation, it would suggest that the stat really does mean something like luck. If there's a strong correlation, it would suggest that it's merely picking up something that's missing from the model.Natenoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-55153306139505304672009-12-09T19:30:37.329-05:002009-12-09T19:30:37.329-05:00By several measures, the Washington Nationals were...By several measures, the Washington Nationals were an extremely unlucky team this past year. I think DC is screwed. I'm glad I live in Seattle now...Jeff Clarkenoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-42448637970911630212009-12-09T18:03:26.475-05:002009-12-09T18:03:26.475-05:00Detroit rank 14th. I don't think they've b...Detroit rank 14th. I don't think they've been so high up a list ever. Still, at 2-10 I think as a fan you'd rather be able to say "we're unlucky". As it is, Detroit fans just have to face reality - they are plain baaaaad :)Iannoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-80041569732295627732009-12-09T17:06:47.309-05:002009-12-09T17:06:47.309-05:00Is the GWP for each team adjusted to their SOS? A ...Is the GWP for each team adjusted to their SOS? A team like the Vikings may not be as lucky if their .61 GWP would be ~.70 based on their easy schedule.Zachnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-17685198799569343372009-12-09T15:38:57.045-05:002009-12-09T15:38:57.045-05:00What would be the difference if luck were measured...What would be the difference if luck were measured by percent change in wins vs wins expected.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-51041253002400557482009-12-09T11:58:43.905-05:002009-12-09T11:58:43.905-05:00Cool post, Brian. Agree with these findings. Worth...Cool post, Brian. Agree with these findings. Worth remembering that while the Saints and Colts are two of the luckiest four teams, they're still the top two teams in expected wins. So they should be the best two teams, and they are, but they just should have more/some losses.Chasehttp://www.pro-football-reference.comnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-30367655783401197782009-12-09T10:29:48.058-05:002009-12-09T10:29:48.058-05:00Your brother in law has an interesting point. Is ...Your brother in law has an interesting point. Is there an easy-ish way to see if anyone has been on a lot of lucky (or unlucky) teams over his career? Maybe that would point to something missing in the model.Alexnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-68831092971951116062009-12-09T10:11:32.671-05:002009-12-09T10:11:32.671-05:00Since the expected wins regression is based on ave...Since the expected wins regression is based on average offensive and defensive performance, whereas actual wins are largely a matter of timing as you point out, would it be more accurate to call the luckiest teams "most clutch?" Or at least, luckiest and most clutch? In the Saints/Skins example, the missed FG was pure luck, but the TD pass to tie it was the perfect outcome at the perfect time.<br /><br />As the Favre case illustrates, WPA accounts for timing as well as overall achievement. Taking a knee on the 1 yard line can give a player a very high WPA relative to scoring the TD, but also result in a large residual against the model, which is based largely on scoring (correct me if I'm wrong here). Maybe WPA can help tease apart luck from timing?Willhttps://www.blogger.com/profile/02178230449052059046noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-23380617294841451912009-12-09T09:57:51.126-05:002009-12-09T09:57:51.126-05:00Steelers fan and I were talking about his team yes...Steelers fan and I were talking about his team yesterday. Of their 6 losses, two have come in the last few seconds of regulation and three in OT. That team probably "should" have 8-9 wins.Jasonhttps://www.blogger.com/profile/09834181305584355651noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-1077634557644544742009-12-09T09:09:05.610-05:002009-12-09T09:09:05.610-05:00dear sweet lord, the redskins are awful. but at le...dear sweet lord, the redskins are awful. but at least now we have an excuse - it's just bad luck!Another Davidhttps://www.blogger.com/profile/07555722546982081280noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-40834542655898147912009-12-09T08:41:18.709-05:002009-12-09T08:41:18.709-05:00"Maybe my brother-in-law is right."
I&#..."Maybe my brother-in-law is right."<br /><br />I'm glad you have developed a statistical model to prove this.The other Briannoreply@blogger.com