Weekly game probabilities are available now at the nytimes.com Fifth Down. This week I also look at how important the various facets of the game appear in the playoffs.Note: One thing I neglected to mention is that the game probabilities this week do take into consideration team performance in the wildcard games.
Seems like you have a few critical responses this week. People just don't understand. Keep up the good work!
ReplyDeleteThis comment has been removed by the author.
ReplyDeleteWant to know something neat? There are eight teams left in the playoffs, with one team from each division!
ReplyDeleteThat's really odd. I just threw the theory of "run D wins championships" into the pile of dreck the "analysts" spout, such as "establish the run. "I'd like to see more on regular season v. postseason in the future!
ReplyDeleteAlso, yeah there seems to be some very negative comments there. Seems they don't actually look at your system's track record vs conventional "wisdom."
ReplyDeleteHas anyone gone back to see how it's done as a whole this season? Did it predict the number of wins it was correct on, and was it accurate (ie .6 probability teams won 60% of the time, .8 teams won 80%, etc)?
ReplyDelete72.1% going into the final week of the season, which I never count. The model was still pretty good that week (10-6) in case anyone thinks I'm cherry picking.
ReplyDeleteHere is a comment at fifth down where a reader looks at the calibration.
http://fifthdown.blogs.nytimes.com/2009/12/31/week-17-nfl-game-probabilities/#comment3
Looks like the model was a little overconfident at the mismatch extremes. But there are relatively few of those games, so it could be just random error.
Any chance we could get updated Super Bowl probabilities?
ReplyDeleteLOL.
ReplyDeleteLook at all of the commentors falling for the error of recency.
As I attempted to say at the NYT site, but apparently it didn't please them:
ReplyDeleteIs it possible that offenses are much more conservative with their playcalling, calling many more runs in the playoffs because the coaches' jobs are much more on the line, and therefore run defense is more important because it gets used more often?
I agree. I've suspected that for a couple years, but it's really hard to back up. It's difficult to measure conservatism in such a small sample. I've wondered if that was always Schottenheimer's Achilles heal in the playoffs. Probably just bad luck, but it could be something more.
ReplyDeleteFWIW, I tracked your probabilities against Vegas lines, and your predictions were slightly better, with a 5.88 RMSE compared to 5.93 for Vegas. Pretty impressive.
ReplyDeleteBut then, come on: something HAS to be wrong if you've got Arizona at only 19%. They have a sixth sense, don't forget.
I reported cumulative numbers for some reason. 0.20 for you and 0.21 for Vegas, on average. NOT statistically significant, unsurprisingly. Still cool though.
ReplyDeleteSome situational factors to consider:
ReplyDelete1) 6 nights of rest = very bad thing for Baltimore and Arizona, especially Arizona who played an OT game with lots of possessions.
2) 8 nights of rest = very good thing for the Jets and Cowboys (mitigates the bye week advantage for their opponents).
3) Good teams that struggle heading into a bye week usually return to form following the bye. This is the case for the Saints, Vikings, and Colts, so I would expect all 3 teams to return to mid-season form.
4) Teams that are on a hot streak heading into a bye week tend to struggle after the bye (overconfidence maybe?). Sorry Chargers fans.
The teams that seem to benefit most from situational factors this week are the Jets, Colts, and Saints.
How would the numbers look if you took the co-efficients and stuff out of the play-off data and ran the probabilities again?
ReplyDeleteIt's possible, but the data gets really thin. When I go back prior to 2002, I don't have penalty data.
ReplyDeleteRegarding run defense becoming more important during the playoffs, I dunno, this may be irrelevant for football, but...
ReplyDeleteThe people at Baseball Prospectus, when examining Billy Beane's lament that 'his stuff doesn't work in the playoffs' to see if it is really true, much to their surprise found data indicating it is, and that the reason is defense becomes significantly more important in the playoffs. In post-season play defensive quality contributes a lot more to victories than one would expect from looking at the larger pool of full season data and what predicts W-L there -- and this has been so for many years, it is not a recent thing.
I forget many of the compelling details, but there's a chapter about it in Baseball Between the Numbers, IIRC. Maybe there is something here that goes across sports. FWIW.
I've always thought that psychology plays a major role in the apparent defensive advantage in playoff games.
ReplyDeleteGood offensive pass the ball. And passing the ball means taking risks. Apparently we as humans get very risk averse in pressure situations.
Have you checked to see if this trend is weather related? The playoffs happen at the coldest part of the year. I don't know the extent to which that really affects passing efficiency, but it may be enough to convince coaches to call more runs, thereby putting a premium on run defense.
ReplyDeleteI bet the people that ridiculed the NOvARI game, are kicking themselves in the a$$ right now.
ReplyDelete