Grist For The Mill

[Deus Irae]

Prior to the New England Patriots first offensive play from scrimmage, the score was 0-0.

That's irrelevant to the point made by BradyFTW!, as you know. Also, crying because the first play of the game for the Patriots was a pass instead of a run is idiotic, to put it kindly. If they'd run and been stuffed for a safety, you'd be screaming that they should have passed.

 

[bucky]

Super Bowl 36 ~ Pass/Run = 27/24 ~ Running = 47% ~ Victory!!

Super Bowl 38 ~ Pass/Run = 48/35 ~ Running = 42% ~ Victory!!

Super Bowl 39 ~ Pass/Run = 33/28 ~ Running = 46% ~ Victory!!

Super Bowl 42 ~ Pass/Run = 48/16 ~ Running = 25% ~ Defeat!!

Super Bowl 46 ~ Pass/Run = 43/19 ~ Running = 31% ~ Defeat!!

Yeah, no correlation, there!!

Great statistical analysis. Very relevant! Here's another one for you:

Super Bowl 36 ~ played in New Orleans = South Region ~ Victory!! :rocker:

Super Bowl 38 ~ played in Houston = South Region ~ Victory!! :rocker:

Super Bowl 39 ~ played in Jacksonville = South Region ~ Victory!! :rocker:

Super Bowl 42 ~ played in Arizona= West Region~ Defeat!!

Super Bowl 46 ~ played in Indianapolis = Midwest Region ~ Defeat!!

Yeah, no correlation, there!!

 

[billdog3484]

I think that the Pats do a very good job drafting overall. Trying to break it down to position and round will just run you into all of the standard problems that you get when you look at a too-small sample size.

oh my goodness. i never said they didnt do a good job in general. i said theyve struggled at drafting WR and DBs in the 2nd and 3rd round. thats it

 

[PatsFanSince74]

Great statistical analysis. Very relevant! Here's another one for you:

Super Bowl 36 ~ played in New Orleans = South Region ~ Victory!! :rocker:

Super Bowl 38 ~ played in Houston = South Region ~ Victory!! :rocker:

Super Bowl 39 ~ played in Jacksonville = South Region ~ Victory!! :rocker:

Super Bowl 42 ~ played in Arizona= West Region~ Defeat!!

Super Bowl 46 ~ played in Indianapolis = Midwest Region ~ Defeat!!

Yeah, no correlation, there!!

it does rely, admittedly, on a limited number of data points, but the OP's analysis does suggest a measurable correlation between play selection and victory; there are no doubt other variables that would have to be considered to get to a compelling level of explanation, but he isn't exactly linking it to a butterfly flapping its wings in China.

 

[Snake Eyes]

Let's run the first offensive play from scrimmage with an empty backfield, so the New York Giants defense don't have to respect a run play whatsoever. In the meantime, the New York Giants defense was ranked 19th against the rush.

I absolutely LOATHE empty backfield plays for this exact reason. Military strategists since antiquity have stressed the importance of the element of surprise, why would you throw that away by telling the defense what you're going to do?

 

[BradyFTW!]

oh my goodness. i never said they didnt do a good job in general. i said theyve struggled at drafting WR and DBs in the 2nd and 3rd round. thats it

Exactly, and I said that, by breaking it down into both position and round, you're dealing with a sample size so small that it's near-impossible to draw broad conclusions from it. It's all right there in my post.

If even one of those misses had hit, your entire conclusion would be completely changed. That fact alone makes it impossible to form a meaningful point.

 

[BradyFTW!]

it does rely, admittedly, on a limited number of data points, but the OP's analysis does suggest a measurable correlation between play selection and victory; there are no doubt other variables that would have to be considered to get to a compelling level of explanation, but he isn't exactly linking it to a butterfly flapping its wings in China.

Actually, it really doesn't suggest very much at all. This statement gets used *way* too often (and frequently incorrectly) on this site, but this is a classic case of correlation--and loose correlation, at that--being mistaken for causation.

On 2001, 2003, and 2004, we had a QB in his early-mid 20s, and we won! in 2007 and 2011, we had a QB who was 30+, and we lost! Clearly, if we ever want to win another SB, we need to trade Brady and start Mallett.

 

[PatsFanSince74]

Actually, it really doesn't suggest very much at all. This statement gets used *way* too often (and frequently incorrectly) on this site, but this is a classic case of correlation--and loose correlation, at that--being mistaken for causation.

On 2001, 2003, and 2004, we had a QB in his early-mid 20s, and we won! in 2007 and 2011, we had a QB who was 30+, and we lost! Clearly, if we ever want to win another SB, we need to trade Brady and start Mallett.

I acknowledged in my comment that there were a limited number of data points and that other variables would have to be introduced to arrive at a satisfactory explanation. That said, correlation is, by definition, always a matter of the degree of correlation and correlation itself is seldom the same as causation, even if the correlation is 1. (The example of Brady v Mallett is, as you intended it, uninteresting statistically, so I won't even address it.)

However, the correlation between play selection and outcomes is not completely random in this case. In fact, with the qualifications that I mentioned in my original post and which I repeated above, if you make Winning and Losing the dependent variable and you make Run/Pass Play Selection the independent variable, the R-square of a regression of those data is north of 0.9, suggesting that the correlation is not "loose," but, at the very least, "observable."

As someone who spends a lot of time with statistics, I hasten to add again, as I have already stated twice, that the limited number of data points make it impossible to draw any final conclusions from the analysis, but one conclusion you cannot draw is that the analysis suggests nothing.

What the analysis suggests is that the exercise is incomplete and that, in the end, the dominant explanatory variable could be completely different than play seleciton, but that play selection will be meaningful.

 

[Deus Irae]

I acknowledged in my comment that there were a limited number of data points and that other variables would have to be introduced to arrive at a satisfactory explanation. That said, correlation is, by definition, always a matter of the degree of correlation and correlation itself is seldom the same as causation, even if the correlation is 1. (The example of Brady v Mallett is, as you intended it, uninteresting statistically, so I won't even address it.)

However, the correlation between play selection and outcomes is not completely random in this case. In fact, with the qualifications that I mentioned in my original post and which I repeated above, if you make Winning and Losing the dependent variable and you make Run/Pass Play Selection the independent variable, the R-square of a regression of those data is north of 0.9, suggesting that the correlation is not "loose," but, at the very least, "observable."

As someone who spends a lot of time with statistics, I hasten to add again, as I have already stated twice, that the limited number of data points make it impossible to draw any final conclusions from the analysis, but one conclusion you cannot draw is that the analysis suggests nothing.

What the analysis suggests is that the exercise is incomplete and that, in the end, the dominant explanatory variable could be completely different than play seleciton, but that play selection will be meaningful.

1.) In 2001, 3003 and 2004, the Patriots had all blacks as their top wide receivers, and they won.

In 2007 and 2011, the Patriots had a white guy in their top wide receiving corps, and they lost.


2.) In 2001, 2003 and 2004, the Patriots had a player named Matt Chatham on the team, and they won.

In 2007 and 2011, the Patriots did not have a player named Matt Chatham on the team, and they lost.


3.) In 2001, 2003 and 2004, the Patriots had a player named Patrick Pass on the team, and they won.

In 2007, the Patriots did not have a player named Patrick Pass on the team, and they lost.


4.) In 2001, 2003 and 2004, the Patriots had Eric Mangini on their coaching staff and they won.

In 2007 and 2011, the Patriots did not have Eric Mangini on their coaching staff, and they lost.



Correlations without context are generally meaningless, especially when the sample size is a small one.

 

[ATippett56]

That's irrelevant to the point made by BradyFTW!, as you know. Also, crying because the first play of the game for the Patriots was a pass instead of a run is idiotic, to put it kindly. If they'd run and been stuffed for a safety, you'd be screaming that they should have passed.

If, If, If. You only supply conjecture. The New York Giants had the 19th ranked rush defense in the NFL.

 

[BradyFTW!]

If, If, If. You only supply conjecture. The New York Giants had the 19th ranked rush defense in the NFL.

And the 29th ranked pass defense.

 

[Deus Irae]

If, If, If. You only supply conjecture. The New York Giants had the 19th ranked rush defense in the NFL.

And the 29th ranked pass defense.

BradyFTW! made a post, you made a really stupid reply to it. I noted an easy-to-see problem with your reply, you went off on an irrelevancy. I responded by pointing out that it was an irrelevancy, and then demonstrated the ridiculousness of your irrelevant post by noting the escape it left you. You then added another irrelevancy, and BradyFTW! has now just clowned you on that.

You should really just either admit you were wrong or walk away, before you look even more foolish.

 

[PatsFanSince74]

1.) In 2001, 3003 and 2004, the Patriots had all blacks as their top wide receivers, and they won.

In 2007 and 2011, the Patriots had a white guy in their top wide receiving corps, and they lost.


2.) In 2001, 2003 and 2004, the Patriots had a player named Matt Chatham on the team, and they won.

In 2007 and 2011, the Patriots did not have a player named Matt Chatham on the team, and they lost.


3.) In 2001, 2003 and 2004, the Patriots had a player named Patrick Pass on the team, and they won.

In 2007, the Patriots did not have a player named Patrick Pass on the team, and they lost.


4.) In 2001, 2003 and 2004, the Patriots had Eric Mangini on their coaching staff and they won.

In 2007 and 2011, the Patriots did not have Eric Mangini on their coaching staff, and they lost.



Correlations without context are generally meaningless, especially when the sample size is a small one.

Correlations in isolation are meaningless. Agreed. And, I noted the small size of the sample. My only point was that the correlation between play selection and outcomes was not insignificant and cannot be dismissed out of hand. Who knows how it would stand up to an analysis of the last 11 seasons as part of a study that looked at 20 or 30 variables? Might end up important; might be not so important.

Cute points on Mangenius and the white guys; coming from anyone but you, I'd take them as a cheap shot. You forgot to indicate whether there were any changes in the Cheerleader outfits; that could have been significant too. :singing:

 

[ATippett56]

And the 29th ranked pass defense.

The New England Patriots offense had following options in the Super Bowl:

Either run the football with healthy BenJarvus Green-Ellis or throw the football to a hobbled tight end with one good ankle. I choose the former.

 

[Deus Irae]

The New England Patriots offense had following options in the Super Bowl:

Either run the football with healthy BenJarvus Green-Ellis or throw the football to a hobbled tight end with one good ankle. I choose the former.

The Patriots had only those 2 options? They couldn't have run the ball with another running back, or thrown the ball to another player?

Were there some laws passed concerning this that only you are aware of?

 

[ATippett56]

The Patriots had only those 2 options? They couldn't have run the ball with another running back, or thrown the ball to another player?

BenJarvus Green-Ellis versus the Baltimore Ravens defense in the AFC Championship Game:

15 rushes for 68 rushing yards which equals to 4.5 yards per carry against the second ranked rush defense in the NFL.

Someone forgot to feed the beast in the Super Bowl against the 19th ranked rush defense in the NFL! Also, Green-Ellis is the type of running back that will wear down an opposing defense, especially against an undersized defensive front.

Were there some laws passed concerning this that only you are aware of?

Ask Belichick why Stevan Ridley was active for the Super Bowl but did not play.

 

[BradyFTW!]

The New England Patriots offense had following options in the Super Bowl:

Either run the football with healthy BenJarvus Green-Ellis or throw the football to a hobbled tight end with one good ankle. I choose the former.

Gronk was only targeted 2-3 times, so no, that wasn't the only other option. There were plenty of other options. Turns out that the Patriots have more than one receiver.

 

[Deus Irae]

BenJarvus Green-Ellis versus the Baltimore Ravens defense in the AFC Championship Game:

15 rushes for 68 rushing yards which equals to 4.5 yards per carry against the second ranked rush defense in the NFL.

Someone forgot to feed the beast in the Super Bowl against the 19th ranked rush defense in the NFL! Also, Green-Ellis is the type of running back that will wear down an opposing defense, especially against an undersized defensive front.

Ask Belichick why Stevan Ridley was active for the Super Bowl but did not play.

So you're just going to keep moving the goalposts rather than admitting that your argument continues to suck....?

 

[BradyFTW!]

I acknowledged in my comment that there were a limited number of data points and that other variables would have to be introduced to arrive at a satisfactory explanation. That said, correlation is, by definition, always a matter of the degree of correlation and correlation itself is seldom the same as causation, even if the correlation is 1. (The example of Brady v Mallett is, as you intended it, uninteresting statistically, so I won't even address it.)

However, the correlation between play selection and outcomes is not completely random in this case. In fact, with the qualifications that I mentioned in my original post and which I repeated above, if you make Winning and Losing the dependent variable and you make Run/Pass Play Selection the independent variable, the R-square of a regression of those data is north of 0.9, suggesting that the correlation is not "loose," but, at the very least, "observable."

As someone who spends a lot of time with statistics, I hasten to add again, as I have already stated twice, that the limited number of data points make it impossible to draw any final conclusions from the analysis, but one conclusion you cannot draw is that the analysis suggests nothing.

What the analysis suggests is that the exercise is incomplete and that, in the end, the dominant explanatory variable could be completely different than play seleciton, but that play selection will be meaningful.

The correlation between QB age and winning Super Bowls isn't random, either. Super Bowls are almost never won by QBs in their mid-30s. But that's irrelevant to this situation, just like OTG's example.

 

[BradyFTW!]

Correlations in isolation are meaningless. Agreed. And, I noted the small size of the sample. My only point was that the correlation between play selection and outcomes was not insignificant and cannot be dismissed out of hand. Who knows how it would stand up to an analysis of the last 11 seasons as part of a study that looked at 20 or 30 variables? Might end up important; might be not so important.

Cute points on Mangenius and the white guys; coming from anyone but you, I'd take them as a cheap shot. You forgot to indicate whether there were any changes in the Cheerleader outfits; that could have been significant too. :singing:

The bold portion is not self-evident. You're stating that it is, but there is still a burden on whoever is arguing this point to explain why A leads to B. Without that explanation, it's nothing more than meaningless, weak correlation. That's the point that DI and I have been making with all of these counter-examples.