1/26: New Charts from Sharp Analysis

[Discotrish]

Interesting chart, but now I'd like to see all other teams broken out. If there are other teams that are similarly outlying, that would throw doubt on the accusation.

The only teams that even come close are the dome teams, that aren't affected by the weather half the time. But New England's fumble rates are a lot better than the dome teams from 2007 on.

Patti

[Mr. Miyagi]

I, for one, find the science and the charts baffling. Deciphering them takes too long and my attention wanders as I lose focus.

Taking a hard look at these though,

Really? It's not that difficult. Just look at that last chart, that's all you need to understand.

What do you do for a living?

[Meathead]

Interesting chart, but now I'd like to see all other teams broken out. If there are other teams that are similarly outlying, that would throw doubt on the accusation.

its based on the same data as the other charts, which showed there was no other team that was similarly outlying. just the cheaters. thats it

[stuckincincy]

I don't understand the bar graphs cited by the OP.

The first one is for 8 "non-dome" clubs. The 2nd is for 10, the 3rd is for 12.

There are but 8 NFL clubs that have domed/retractable roof stadiums. Shouldn't these charts depict 24 clubs? Are they "cherry picked"? Are there other clubs close to NE?

I'm certainly not defending NE, but I am reminded of the old adage..."If you torture numbers long enough, they will admit to anything."


http://en.wikipedia.org/wiki/List_of...eague_stadiums

[Meathead]

i assumed he was just taking a sample of teams in the normal spectrum so the chart wouldnt be freaking huge. doesnt really need to represent all teams since he has that data elsewhere in other graphs

and it is true that you can make stats say just about anything you want, but you have to use subterfuge to do that. this data is pretty straight forward i think - number of total fumbles. and the difference is so dramatic that its hard to imagine you could finagle a lie that big if the cheaters were actually close to the pack over that period

i did see an article someone else wrong trying to debunk sharps analysis but i thought it was rather poorly done and focused mostly on petty things that even if valid would not account for the huge disparity the cheaters enjoyed. ill see if i can find that article, it was just yesterday

[stuckincincy]

i assumed he was just taking a sample of teams in the normal spectrum so the chart wouldnt be freaking huge. doesn't really need to represent all teams since he has that data elsewhere in other graphs

and it is true that you can make stats say just about anything you want, but you have to use subterfuge to do that. this data is pretty straight forward i think - number of total fumbles. and the difference is so dramatic that its hard to imagine you could finagle a lie that big if the cheaters were actually close to the pack over that period

i did see an article someone else wrong trying to debunk sharps analysis but i thought it was rather poorly done and focused mostly on petty things that even if valid would not account for the huge disparity the cheaters enjoyed. ill see if i can find that article, it was just yesterday

Nah - the presentation really doesn't pass muster. It reminded me of Al Gore's charts in his political hack flick, "An Inconvenient Truth.", which garnered 2 Oscars and the awarding of a Nobel Peace prize for zinc mining, yacht-owning, multiple mansion-owning Al.

It's really unacceptable science - smells of an agenda - to reduce the population of a data set. Of course, governments do that as a matter of course all the time. And I'm wondering now if the referenced site is trying to now cover their fannies.

[Strongman]

I don't understand the bar graphs cited by the OP.

The first one is for 8 "non-dome" clubs. The 2nd is for 10, the 3rd is for 12.

There are but 8 NFL clubs that have domed/retractable roof stadiums. Shouldn't these charts depict 24 clubs? Are they "cherry picked"? Are there other clubs close to NE?

I'm certainly not defending NE, but I am reminded of the old adage..."If you torture numbers long enough, they will admit to anything."


http://en.wikipedia.org/wiki/List_of...eague_stadiums

It's a histogram used to show frequency. In those graphs, each bar doesn't represent a team. A bar is a representation of fumbles per play, while the bar height shows how many teams this happened to.

Think of it look this: If you have 24 coins (teams) and flipped them all at once and counted the results (heads or tails). For illustration's sake we'll say 14 coins were heads and 10 coins were tails. You would make a histogram with two bars and the bars would be 14 and 10 units high.

[Strongman]

Nah - the presentation really doesn't pass muster. It reminded me of Al Gore's charts in his political hack flick, "An Inconvenient Truth.", which garnered 2 Oscars and the awarding of a Nobel Peace prize for zinc mining, yacht-owning, multiple mansion-owning Al.

It's really unacceptable science - smells of an agenda - to reduce the population of a data set. Of course, governments do that as a matter of course all the time. And I'm wondering now if the referenced site is trying to now cover their fannies.

Histograms are used in science all of the time. They are a valid way to illustrate the frequency of an occurence.

[gebobs]

I don't understand the bar graphs cited by the OP.

The first one is for 8 "non-dome" clubs. The 2nd is for 10, the 3rd is for 12.

Each bar shows a frequency i.e. how many teams were in each "bucket" of fumble rates. The first shows buckets of 4,4,2,4,3,4,1,1...totaling 23 teams. The second show 1,1,2,4,4,4,3,2,1,1...totaling 23 teams. The third shows one superimposed on the other.

[harmonkoz]

What I find interesting is all of the people who came out last week debunking Sharp. There were some good write-ups about how wrong Sharp was.

http://drewfustin.com/2015/01/27/pat...mble-comments/
http://regressing.deadspin.com/why-t...mos-1681805710

There were many, many more. Almost a concerted push back.

I kind of forgot about it, but checked back in on Mr. Sharp today. http://www.sharpfootballanalysis.com...ate-since-2000

Needless to say, he is sticking to his guns.

[gebobs]

http://drewfustin.com/2015/01/27/pat...mble-comments/

Sure, there are lots to take umbrage with in Sharp's analysis, but this one left me scratching my head...

"The Patriots are indeed nearly off the chart, but that is partially because the author uses the smallest y-axis possible to demonstrate the largest effect that he could. It's generally preferred to use a y-axis that begins at 0..."

I read this objection elsewhere and it's nonsense. In fact, it's generally preferable to use a tight axis in order to show the data with precision. There is no general rule for a y-axis origin at zero just as there is none for a y-axis maximum of 100. It's completely arbitrary. It's generally assumed that the reader is going to understand the scale. Regardless, changing that y-axis origin to zero wouldn't materially effect how the chart looks.