SalishHawkFan
New member
- Joined
- Apr 30, 2009
- Messages
- 5,872
- Reaction score
- 0
Drew Fustin a PhD from Chicago, who is right up front admitting he's a Patriots fan, goes into some detail debunking Warren Sharps data that shows the Patriots fumble less than any other team from 2007 onwards.
I just left him hoisted on his own petard by showing his own formula proves that the Patriots odds of doing what they did are 1.4 million to 1.
There's math, but hang in there 12's, we're the most literate people in the nation:
http://drewfustin.com/2015/01/27/patrio ... -comments/
First he challenges some of Sharps assumptions. Now he's totally on the mark in his challenges, that's what makes his final downfall so very sweet.
Challenge #1: Don't throw out domed teams.
Sharp throws out domed teams and looks at just teams that play outdoors because domed teams don't fumble as often. Fustin correctly points out that you shouldn't throw out domed teams in games where they played outdoors.
This is a very valid challenge. The data that he then gives is compiled using domed teams while playing outdoors. Better data than Sharp used.
Challenge #2: Don't cherry pick a rolling 5 year average. Totally on the mark here. in 2013 the Patriots had a LOT of fumbles and ranked 21st. Using anything but a rolling 5 year average doesn't make the Pats look so bad. Sharp shouldn't have done that.
As I will soon show, Fustin just drew his own petard. I will soon leave him hoisted upon it.
Challenge #3: Don't say this can only occur once in 16,233.77 instances
Now he points out another mistake that Sharp made. It's only a mistake, however, because of challenges 1 and 2. Using correct data by including domed teams that played outdoors, that number is going to change. He then gives us the formula to calculate what the actual % chance is that the Patriots do as well as they've done. But it's not an apples to apples comparison. He decides to show us instead what the odds are of the Pats finishing in the top 3 in least fumbles are in a 5 year period.
Umm, Dr. Fustin, didn't you just say that the rolling 5 year average was flawed?
Petard, prepare to hoist.
He gives us his formula and calculates that finishing in the top 3 4 times in 5 years has only a .004% chance of happening, which comes to 2800 to 1 odds. That's based upon merely a 3/32 chance of finishing top 3. It doesn't include skill or scheme. So a team that really stresses hanging onto the ball and calls plays that are less likely to result in turnovers and only hires players who are less likely to fumble will do better.
All of which is true.
Here's his data, adjusted correctly for domed teams playing outdoors.
2007-present NE ranking for plays per fumble: 2nd, 5th, 2nd, 1st, 2nd, 3rd, 21st and 2nd
2002-2--6 NE ranking for plays per fumble: 14th, 10th, 18th, 10th, 19th
Now, fine, the odds of finishing top 3 four out of five years is about 2800 to one. But Dr. Fustin just said using rolling five year averages was cherry picking didn't he? So let's use his words of wisdom against him shall we?
HOIST AWAY!!!!
Using his own formula, what are the odds that NE can finish top 3 six out of eight years as his very own data shows?
p=a^8+8a^6(1-a) where a=3/32.
p=.000007% or 1.4 million to 1 odds.
Sorry, doc, don't try to bait and switch us Seahawk fans, we're too sharp for those kinds of con games. By his own admission, NE finished top 3 six of eight seasons and by his own formula, the odds of that are 1.4 million to 1 if you don't calculate for skill and scheme, just luck.
Only problem is, he himself points out that NE doesn't hire players less likely to fumble, in fact, he shows they are 23% MORE likely to fumble when they play for other teams than when they play for NE.
So strike skill from the reason NE always ranks so high. Scheme did not change drastically from 2006 to 2007.
No, you cannot adjust for skill and scheme cannot make up the difference because scheme has changed only slightly. That just leaves cheating.
I just left him hoisted on his own petard by showing his own formula proves that the Patriots odds of doing what they did are 1.4 million to 1.
There's math, but hang in there 12's, we're the most literate people in the nation:
http://drewfustin.com/2015/01/27/patrio ... -comments/
First he challenges some of Sharps assumptions. Now he's totally on the mark in his challenges, that's what makes his final downfall so very sweet.
Challenge #1: Don't throw out domed teams.
Sharp throws out domed teams and looks at just teams that play outdoors because domed teams don't fumble as often. Fustin correctly points out that you shouldn't throw out domed teams in games where they played outdoors.
This is a very valid challenge. The data that he then gives is compiled using domed teams while playing outdoors. Better data than Sharp used.
Challenge #2: Don't cherry pick a rolling 5 year average. Totally on the mark here. in 2013 the Patriots had a LOT of fumbles and ranked 21st. Using anything but a rolling 5 year average doesn't make the Pats look so bad. Sharp shouldn't have done that.
As I will soon show, Fustin just drew his own petard. I will soon leave him hoisted upon it.
Challenge #3: Don't say this can only occur once in 16,233.77 instances
Now he points out another mistake that Sharp made. It's only a mistake, however, because of challenges 1 and 2. Using correct data by including domed teams that played outdoors, that number is going to change. He then gives us the formula to calculate what the actual % chance is that the Patriots do as well as they've done. But it's not an apples to apples comparison. He decides to show us instead what the odds are of the Pats finishing in the top 3 in least fumbles are in a 5 year period.
Umm, Dr. Fustin, didn't you just say that the rolling 5 year average was flawed?
Petard, prepare to hoist.
He gives us his formula and calculates that finishing in the top 3 4 times in 5 years has only a .004% chance of happening, which comes to 2800 to 1 odds. That's based upon merely a 3/32 chance of finishing top 3. It doesn't include skill or scheme. So a team that really stresses hanging onto the ball and calls plays that are less likely to result in turnovers and only hires players who are less likely to fumble will do better.
All of which is true.
Here's his data, adjusted correctly for domed teams playing outdoors.
2007-present NE ranking for plays per fumble: 2nd, 5th, 2nd, 1st, 2nd, 3rd, 21st and 2nd
2002-2--6 NE ranking for plays per fumble: 14th, 10th, 18th, 10th, 19th
Now, fine, the odds of finishing top 3 four out of five years is about 2800 to one. But Dr. Fustin just said using rolling five year averages was cherry picking didn't he? So let's use his words of wisdom against him shall we?
HOIST AWAY!!!!
Using his own formula, what are the odds that NE can finish top 3 six out of eight years as his very own data shows?
p=a^8+8a^6(1-a) where a=3/32.
p=.000007% or 1.4 million to 1 odds.
Sorry, doc, don't try to bait and switch us Seahawk fans, we're too sharp for those kinds of con games. By his own admission, NE finished top 3 six of eight seasons and by his own formula, the odds of that are 1.4 million to 1 if you don't calculate for skill and scheme, just luck.
Only problem is, he himself points out that NE doesn't hire players less likely to fumble, in fact, he shows they are 23% MORE likely to fumble when they play for other teams than when they play for NE.
So strike skill from the reason NE always ranks so high. Scheme did not change drastically from 2006 to 2007.
No, you cannot adjust for skill and scheme cannot make up the difference because scheme has changed only slightly. That just leaves cheating.