Patriots Ball Security Study

[Urgent]

...please merge into one of the many science threads if this has been mentioned elsewhere, but thought it deserved it's own discussion.

Warren Sharp, at Sharp Football Analysis, has done a study of fumble rates, and shown how the Patriots have improved at this.

http://www.sharpfootballanalysis.co...ts-prevention-of-fumbles-is-nearly-impossible

From 2000 to 2006, he reports, the Patriots were in the middle of league in terms of fumbles. From 2007 to 2014, they have been the best team in the league at avoiding fumbles.

Unfortunately, Sharp titles his analysis "The New England Patriots prevention of fumbles is nearly impossible. His point, I guess, is that, within a sample, it is statistically nearly impossible to vary from the mean or have one data point the best and one the worst.

Conspiracy language aside, it points out the success of Belichick's increasing emphasis on ball security. The wet-ball drills, the outdoor practices, the benching for fumbles. A big contributor to the Patriots success.

 

[Brock Samson]

Why Those Statistics About The Patriots' Fumbles Are Mostly Junk: http://regressing.deadspin.com/why-...e-patriots-fumbles-are-mos-1681805710/+kylenw

 

[reflexblue]

I know why the team didn't fumble much and it has nothing to do with air pressure. If you fumbled you didn't play, ask S Ridley.

 

[wolverinejoe80]

I know why the team didn't fumble much and it has nothing to do with air pressure. If you fumbled you didn't play, ask S Ridley.

but haters bring up lawfirm not fumbling in his 500 carries with us. and then goes to other team he fumbles 5 times in 500 carries.

haters will always find something to hate

 

[Darrone]

The biggest statistical errors that this analysis makes are as follows:

• Treats all plays the same: IE - a run play has the same statistical likelihood of a fumble as a pass play - this is false.
• Treats all QB's the same: The two quarterbacks who have the quickest release (TB and at the time of analysis, Peyton manning's colts) also happen to lead teams with the fewest fumbles. This correlation is not mentioned. Also - teams with inconsistent and rotating QBs have the highest fumble rates.
• Does not address fumbles, only lost fumbles: Recovering a fumble is pure dumb luck. If you are drawing a correlation between fumbles and ball pressure, and stat to use is fumbles, not fumbles lost. Ball pressure would not help recover a fumble.
• Argues the patriots record in bad weather is statistically adherent, but ignores the fact that the Patriots record over the time period included is statistically adherent REGARDLESS of weather. IE - the patriots won a lot of games during bad weather ignores the fact that they won a lot of games regardless of weather.
• Patriots only lead the league in run-plays-per fumble, not in pass-plays-per fumble.

 

[Kasmir]

Here's an article from a statistician debunking this contrived "study":

http://regressing.deadspin.com/why-those-statistics-about-the-patriots-fumbles-are-mos-1681805710

 

[Bobsyouruncle]

I agree, the Patriots are awesome.

 

[Bobsyouruncle]

The biggest statistical errors that this analysis makes are as follows:

• Treats all plays the same: IE - a run play has the same statistical likelihood of a fumble as a pass play - this is false.
• Treats all QB's the same: The two quarterbacks who have the quickest release (TB and at the time of analysis, Peyton manning's colts) also happen to lead teams with the fewest fumbles. This correlation is not mentioned. Also - teams with inconsistent and rotating QBs have the highest fumble rates.
• Does not address fumbles, only lost fumbles: Recovering a fumble is pure dumb luck. If you are drawing a correlation between fumbles and ball pressure, and stat to use is fumbles, not fumbles lost. Ball pressure would not help recover a fumble.
• Argues the patriots record in bad weather is statistically adherent, but ignores the fact that the Patriots record over the time period included is statistically adherent REGARDLESS of weather. IE - the patriots won a lot of games during bad weather ignores the fact that they won a lot of games regardless of weather.
• Patriots only lead the league in run-plays-per fumble, not in pass-plays-per fumble.

What are the odds of one statistician making all those errors? I suspect he's using a deflated brain.

 

[Oswlek]

This has been refuted several times over, but I'm sure it's about to become the next hot topic.

 

[JJC]

What are the odds of one statistician making all those errors? I suspect he's using a deflated brain.

Sharp is NOT a statistician, as far as I can tell. His website fails to pimp his actual academic credentials. All we know is that he is a gambler with a system that he is pimping. Right, you pay him for access to his statistical-cannot-miss tips in order to bet on all manner of football minutiae.

As a formally educated engineer with multiple degrees, I believe in physics, chemistry, and calculus, not statistics. Statistics do not tell you definitively how or why something occurs. It only tells you that when A occurs, B occurs with some corresponding regularity. Only real science, supported by hypothesis, experimentation, and theory can tell us HOW or WHY A causes B to occur.

The Patriots have put forth a plausible physics based explanation of what occurred, how it occurred, and challenged the world to prove them wrong. They are on the side of science and truth.

Everyone else is starting with a conclusion and cherry picking observation to support that conclusion. They are the mob.

I have feared for many years that we are on a trajectory toward a new Dark Age, with the apogee of our modern achievements being the Apollo missions. This whole affair seems to support that depressing prediction.

Patriots 45, Seahawks 24

 

[juny]

http://fivethirtyeight.com/datalab/your-guide-to-deflate-gateballghazi-related-statistical-analyses/

this link should kill that flawed study, but it will get 1/10000 the attention

 

[50-yard-line]

The linked article by Kasmir does a good job of poking lots of huge holes - like regarding Brandon Tate. Look how much he fumbled since he left New England!!!! Yeah, duh, returning kicks and punts...

The REAL anomaly here is Ben-Jarvis Green-Ellis. If you've got your prime running back who simply doesn't fumble, well, yeah, you'll be lower than most other teams.

Over 500 carries with New England without a fumble, and before that...920 carries without a fumble in 4 years of college ball.

Over 1400 carries without a fumble! Are you kidding me??????

He fumbled 5 times after that. Age? Reality catching up with him?

Also, New England does focus on ball security more than other teams, and in some ways, it hurts them. I've noticed that Ridley isn't hitting the hole as fast and hard as he used to. He's got his arms up higher, protecting the ball and is spending more thought/strength/etc. clutching that pigskin.

if you look at his fumble in, say, the Denver game last year, he tried to make a nifty spin move with the ball down low and wide. Other teams might accept that lack of ball security because of the extra yards such "reckless" running might net you. Bellichick most certainly will not, as Ridley learned the hard way.

 

[FourierSeries]

Why Those Statistics About The Patriots' Fumbles Are Mostly Junk: http://regressing.deadspin.com/why-those-statistics-about-the-patriots-fumbles-are-mos-1681805710/ kylenw

I knew there was something off about their assumption of normality in terms of plays per fumble. I can see fumbles per play being normally distributed, but theres no reason for the inverse to be normal as well.

Very good article and it completely refutes the claim that they had to have cheated.

Gotta love statistics (when done right)!

 

[FourierSeries]

As a formally educated engineer with multiple degrees, I believe in physics, chemistry, and calculus, not statistics. Statistics do not tell you definitively how or why something occurs. It only tells you that when A occurs, B occurs with some corresponding regularity. Only real science, supported by hypothesis, experimentation, and theory can tell us HOW or WHY A causes B to occur.

I agree with most of this, although I have a bit of a different take.

I trust the use of statistics in the aforementioned areas - physics, chemistry, etc. This is because these fields allow for controlled experiments that can be replicated in order to generate multiple samples of sufficient size. Any statistical model requires high quality data for it to produce useful output. When you deal with random phenomena (which are present in physics and chemistry) it's almost impossible to utilize anything other than statistics & probability to analyze the problem.

The issue is with the misuse of statistics, as portrayed in the study claiming the Pats had to have cheated given their record of fumbling. In addition to the theoretical errors (using the wrong distribution when discussing plays per fumble) the author also failed to utilize high quality data as inputs. Which led to erroneous output.

This problem is common when applying statistics (and mathematics in general) to social science (and sports.) We see it in economics, in finance, and many other fields. For those who aren't aware, the amount of statistics and mathematics used in finance over the years has grown exponentially. They use mathematical concepts from physics and apply them to financial problems like pricing options and risk management. This reliance on stats/math (and more importantly the misuse of it) was a large reason for the financial crisis several years ago.

Anyways, getting way off topic, but it's very bothersome when people misuse statistical analysis and mathematics in our society.

GO PATS!!