f13.net forums - NFL 2014

Quote from: K9 on January 28, 2015, 08:24:28 AM

Quote from: Nevermore on January 27, 2015, 05:42:33 PM

So in 2006, Tom Brady and Peyton Manning lobbied for a rule change that allowed each team to supply their own footballs for use on offense. Prior to that, the home team would supply all the balls used by each team. This article is an interesting analysis of the dramatic change in the rate the Patriots fumbled since that rule was change.

This isn't an awful analysis, but how about Indianapolis and Denver, did they see matching benefits, since Peyton was also lobbying for this? Fumbles are such complex events I don't think trying to boil them down with this sort of simple, one-dimensional analysis really achieves anything useful. The point where they try to calculate a probability hinges on the assumption that fumble rates by team are normally distributed, which seems at best naive and over-simple, at worst completely and utterly wrong. Infrequent events like fumbles and interceptions tend to follow long-tailed distributions; talking about the average rate of fumbles or interceptions by a team or player is nonsensical. Dome teams should have been presented separately as a comparison group; but even more so the analysis should adjust by the number of dome games non-dome teams play on the road. Consider the Panthers or Bucs, they are guaranteed two dome road games a year (NO and ATL), likewise for the Bears and Packers (MIN and DET), whereas NE is guaranteed none. Furthermore it should present the statistics for home and away games, we have no way of assessing the differential that existed before 2007 with what the article presents. Given all this and the complex factors of player skill, variable weather, and differences in coaching; attempting to argue that this boils down to something as trivial as a rule change feels like a real stretch

As Chris Long points out in this article on Deflategate (it's a good read, go on), the Patriots have a tendency to obsess over details like turnovers in a way that other franchises don't. Perhaps they just drafted players with better hands, maybe they figured out better drills for their backs and recievers, maybe they got lucky.

Quote

Let’s get down to it: this story isn’t as much about air pressure as it is about the cult of the New England Patriots. The Patriots are really good at two things: winning football games and not giving a shit what you think about them. This modus operandi has earned the Patriots an equal number of fans and haters.

Unfortunately the 538 article has only one decent point, that he asked the wrong question. The question should be "what is the probability we get at least one team out of the 23 outdoor teams this far away" and not "what is the probability this team is this far away:

The Deadspin article is actively bad, messing up the core epistemology in an article about the core epistemology(in at least 3-4 different ways) and then making broad worthless proclamations about it.

For the record, fumbles are normally distributed however you record them. If you record them as fumbles/play. They're a binomial distribution (which tends to normal for large n, which we have). If you record plays/fumble is the sum of a geometric distribution divided by the trials, which, since geometric distributions are well behaved(have well defined expectation and variance) also tends to normal via the CLT. So you can ignore all the "well its not normal" arguments right there. You can especially ignore "its not normal" arguments that test with the patriots data(I am looking at you fucking Tom Hayden) because outliers(I.E. data pulled from a different distribution) will fucking throw those tests off and turn the test, into essentially, a question of whether or not the outliers are outliers. Bitches don't even fucking know their CLT's.

Using a normal distribution there(if you did not put the NE data in the test) should not give you any qualms as these people suggest. Because we know the form of the event we are recording and we know that it approximates normal after a reasonable amount of trials. Granted the fit isn't as good for plays/fumbles because we have, effectively, far fewer "trials", but we are still talking about 100 fumbles(60/play 6000 plays) over a 5 year period good enough to be reasonably close(though its important to note that the fit gets worse as a teams fumbles/play gets worse since there are fewer total trials). And, importantly, since we know the form of the event we are recording and it can come in no other form not an excuse to discount the results.

Drew Huston(and others) ignore the central thesis in his test because he tests over the entire 2001 to 2014 range rather than the post 2007 range when the teams got a hold of the balls. And then he takes a whole bunch of teams which played in only comparatively few outdoor games (with no accounting for the weather in those games) and compares them to teams which play a lot of outdoor games as if the samples have the same variance. Which they do not. Which means that the pats very good score is still an anomaly.

Cris Long ignores the fact that Belicheck has obsessed over fumbles since forever, but only after the rule change which gave teams control of the footballs did they start to become an anomaly in turnover probability

Sharps analysis was imperfect, but it was far from as flawed as anyone thinks. And they could have simply calculated the probabilities that they wanted if they wanted to really check. But they didn't, and they didn't because when you run the numbers (as I have: though, even I, sigh, have an error which I am too lazy to correct) the results are not flattering to the Patriots. In that the probability that at least one team of 23 indoor would hit so well over a 5 year period(i used the latest since i was cribbing their data) is about 1.5%. These numbers should go down if i used the 2007 to 2014 data.

So yea, the pats ball handling is still a significant anomaly, and we can't attribute it to much aside from the fact that the pats have been deflating their footballs since well, every team does everything else. And the Pats are the only one with so ridiculous numbers when you look at the proper stats.

Quote from: HaemishM on January 27, 2015, 08:10:34 AM

So Robert Kraft says that should the NFL's investigation of Tom Brady's balls not find any evidence of wrongdoing on the part of the Patriots' organization (it won't - and isn't it convenient how the report won't even be concluded until after the Super Bowl), the NFL should issue an apology to Belicheck and Brady.

I got two words for Robert Kraft and they start with the letters Fuck You.

Well Goodel has two words for Robert Kraft and they're "My Friend!"