Couldn't getting accustomed to the softer balls on offense have played a part
in Welker's fumbling the harder balls used for the kicking game?
Ok, we did some stats tests here - that's what I do for a living though I promised myself I'd never do this kind of thing on the board. I had a PhD professor colleague who hasn't watched a football game since the 60's (eg, could care less about the sport) do the analysis. Any of the stats folks can check the work / decide if it holds water. I think this is pretty damning.
"had some fun w/Sharp's graph of offensive runs per fumble in the
attached file. First, I looked at the distributions w/ and w0/187. Note
how "normal" the league distn is wo/187, confirming the notion that fumbles
are simply statistical random acts. Based on this distn I calculated the prob of a team
scoring 187 - WoW! Next, I applied the Grubbs stat outlier test which
retains the 187 in the calculation; again wow! Both cases are damning as
hell. Note that the comments of Prof Greg Matthews denying the Patriots'
outlier contain no analysis. Sharpe's overall analyes (runs/fumble,
2006/2007 break point coinciding w/inflation change, change in players'
fumbles after coming to the P's, the wet vs. dry performance) is brilliant.
Note the very normal distribution of the offensive runs/fumble without the Patriots. Now for this established distribution, the Z-score and probability of some team coming along, performing at the 187 level are:
Z = (187- 102.45)/15.51 = 5.45!!!!! The probability for this kind of performance would be 2.79x10^(-8), in other words P = 0.0000000279 or about 3 in 100 million!
Next, I applied the statistical Grubbs outlier test. This is more stringent than the above discussion in that it retains the questionable value in the distribution, which inflates the std dev (produces a lower P-value). The test does the same Z calcn but w/187 retained, and compares this "Z" to a tabulated "critical" values corresponding to the degs of freedom. If the "Z" comes out above the critical value, then the value is an outlier (>/= 95% confidence) and can be discarded. In this case the value of 3.83 is well above the critical value of 2.94. A corresponding P-value can be calculated, which in this case says that the probability of such performance is about 0.0000043.
I love this Prof of Stat, Greg Matthews comments”
“It’s 98 percent bunk,” said Greg Matthews, an assistant professor of statistics at Loyola University in Chicago. “He basically reached a conclusion already and he wants to find the most sensational stats he can find.”
Matthews said he found a “tremendous amount” of flaws in Sharp’s breakdown, among them the touches-per-fumble method of lining up the Patriots against the rest of the league.
“I refute the fact that the Patriots are an outlier,” Matthews said. “I refute that fact definitely. Are they better after [2006]? Possibly, but it’s not outrageously better.”
Note that he doesn’t refute Sharpe’s calculated difference, he just doesn’t want to put it to a statistical test (of course not; he’s an admitted Patriots’ fan)"
So basically, the Pats won the lotto, got hit 3 times by lightning on there way back home, survived and won the lotto again...Ok, we did some stats tests here - that's what I do for a living though I promised myself I'd never do this kind of thing on the board. I had a PhD professor colleague who hasn't watched a football game since the 60's (eg, could care less about the sport) do the analysis. Any of the stats folks can check the work / decide if it holds water. I think this is pretty damning.
"had some fun w/Sharp's graph of offensive runs per fumble in the
attached file. First, I looked at the distributions w/ and w0/187. Note
how "normal" the league distn is wo/187, confirming the notion that fumbles
are simply statistical random acts. Based on this distn I calculated the prob of a team
scoring 187 - WoW! Next, I applied the Grubbs stat outlier test which
retains the 187 in the calculation; again wow! Both cases are damning as
hell. Note that the comments of Prof Greg Matthews denying the Patriots'
outlier contain no analysis. Sharpe's overall analyes (runs/fumble,
2006/2007 break point coinciding w/inflation change, change in players'
fumbles after coming to the P's, the wet vs. dry performance) is brilliant.
Note the very normal distribution of the offensive runs/fumble without the Patriots. Now for this established distribution, the Z-score and probability of some team coming along, performing at the 187 level are:
Z = (187- 102.45)/15.51 = 5.45!!!!! The probability for this kind of performance would be 2.79x10^(-8), in other words P = 0.0000000279 or about 3 in 100 million!
Next, I applied the statistical Grubbs outlier test. This is more stringent than the above discussion in that it retains the questionable value in the distribution, which inflates the std dev (produces a lower P-value). The test does the same Z calcn but w/187 retained, and compares this "Z" to a tabulated "critical" values corresponding to the degs of freedom. If the "Z" comes out above the critical value, then the value is an outlier (>/= 95% confidence) and can be discarded. In this case the value of 3.83 is well above the critical value of 2.94. A corresponding P-value can be calculated, which in this case says that the probability of such performance is about 0.0000043.
I love this Prof of Stat, Greg Matthews comments”
“It’s 98 percent bunk,” said Greg Matthews, an assistant professor of statistics at Loyola University in Chicago. “He basically reached a conclusion already and he wants to find the most sensational stats he can find.”
Matthews said he found a “tremendous amount” of flaws in Sharp’s breakdown, among them the touches-per-fumble method of lining up the Patriots against the rest of the league.
“I refute the fact that the Patriots are an outlier,” Matthews said. “I refute that fact definitely. Are they better after [2006]? Possibly, but it’s not outrageously better.”
Note that he doesn’t refute Sharpe’s calculated difference, he just doesn’t want to put it to a statistical test (of course not; he’s an admitted Patriots’ fan)"
thats awesome...and yes, that's basically what the data suggests!So basically, the Pats won the lotto, got hit 3 times by lightning on there way back home, survived and won the lotto again...
i think you are right about the outside interested parties thing. The league has to realize they can't hide this stuff as access to data and an analysis of it is too easy now.Thanks royalshank! It's becoming clear to me that it will be very difficult for the NFL to sweep this under the rug. Even putting fans aside, there will be many interested parties looking for answers to the questions this data raises. I think we will soon find out if justice is possible in NFL.
Any one else think Steve Belichick looks a bit sketchy like his dad?
And here's Hines Ward on Late Night with Seth Meyers taking about the Patriots.
[video=youtube;wnKM4feYpt8]https://www.youtube.com/watch?v=wnKM4feYpt8&feature=youtu.be[/video]
And here's Hines Ward on Late Night with Seth Meyers taking about the Patriots.
[video=youtube;wnKM4feYpt8]https://www.youtube.com/watch?v=wnKM4feYpt8&feature=youtu.be[/video]