PowerRun":2nn53xy2 said:
vonstout":2nn53xy2 said:
PowerRun":2nn53xy2 said:
SirTed":2nn53xy2 said:
I don't REALLY care because I don't think it's going to be a factor in the game on Sunday, but I'm not sure if this proves anything, either way. It's circumstantial, but those fumble rate numbers are too staggering to be ignore. Never proved, but they don't look good.
.
Nah. That poorly conducted study was also debunked by someone who actually understands statistics.
The inputs used for the flawed study were flawed themselves. And plays per fumbles is not normally distributed as the author claims.
http://regressing.deadspin.com/why-thos ... 10/+kylenw
Do you know anything about statistics and probabilities? I do. I'm a mechanical engineer. When your cheating team doesn't get to tamper with the balls next year and your turnovers go back to where they were (near the league average), it will prove you were cheating. The guy trying to refute the plays per fumble article is a NE fan. He looks as pathetic as you do trying to defend your team.
I have a degree in mathematics. Mechanical engineers don't do any rigorous stats/probability courses. It's mostly application, which is useful, but doesn't lend to fostering an understanding of the theoretical underpinning of the subject.
Plays per fumbles don't follow a Gaussian distribution. What the other author did was look at fumbles per play, which do follow a Gaussian distribution, and assumed that the inverse is also. But if you know stats like you claim you do, you'd know the inverse need not have the same distribution.
He also uses fumbles lost, and not fumbles. And only uses stats for home games, which is irrelevant.
Try harder.
I'm sure you know Brian Burke. He's one of the ones who ushered in NFL analytics and is the long time administrator of Advanced NFL Stats, developed EPA and WPA, and etc. Here's what he had to say about that Deadspin article:
"This article is so disingenuous it makes stat guys look bad and I don't like it.
http://regressing.deadspin.com/why-thos ... 1681805710 …"
"To be clear, the Regressing article is problematic. Bad numbers, bad model, bad assumptions, bad logic."
"I don't think the fumble numbers prove anything. I also disagree with the idea there's nothing to see here. More than warrants more digging."
Feel free to check his twitter timeline.
https://twitter.com/adv_nfl_stats
So, no one cares if the Pats will be proved to have cheated. The preponderance of the evidence suggests they were, however, whether it can be proved or not.
1. The Pats fumble rate dramatically improved in 2007.
2. In 2007 a rule was instituted allowing offenses to keep control of their footballs.
3. Brady was a lobbyist for this rule.
4. The Pats were caught with underinflated footballs, providing a plausible explanation for how the Pats fumble rate improved.
5. The Pats kick returners and punt returners fumble rates DID NOT improve:
http://imgur.com/3ecTZlM. So, whatever mysterious technology Belichick discovered in 2007, he did not extend to the Pats returners.
Note that the Pats improved fumble rate was only discovered AFTER the underinflated footballs came to light. Someone had a hypothesis that the Pats might have had an improved fumble rate because of the underinflated footballs, and evidence for this hypothesis was found. That makes the evidence very convincing.
You can poke holes in any of these points individually, but together, as a Bayesian matter, an honest person's prior should move toward "Pats cheated".
Now, I don't give a flying leap whether the Pats get sanctioned. That they won't be able to cheat in this game is good enough for me, but that would have been the case anyway, since in the Super Bowl offenses don't get to keep their own footballs[1]. It's probably a moral victory that the Pats won't get to cheat in successive regular seasons, as I'm sure this rule will be changed.
Over the next two years, I expect the Pats fumble rate to revert to pre-2007 levels.
[1] You'll note that post-2007, the Pats offense has underperformed in the Super Bowl, by the way.