Broaddus Tweets: Ratliff will be 3, Hatcher the 1

[theogt]

Exactly what way have you analyzed these correlation? I cant make any specific comments until then.

Correlation between YPA and winning percentage. You can run the analysis in excel.

Off the top of my head, possible reasons why it may SEEM like running is not useful.

It's statistically proven that running better is not useful. It's a knowable, indisputable fact. I'm not sure we need reasons as to why it would "seem" to be true given it's a provable certainty.

In regards to success rate, football is different than most statistical analyses in that EVERY metric is inherently related to winning. Yards, posessions, etc are all innately tied to winning. That is also the reason why making claims like running well doesnt lead to winning makes no sense. Ill give you that success rate is more intrinsically tied to winning than just ypa, but that version of success rate is almost identical to the version that simply predicts probability of first down, which then would be no different than any other metric.

The success rate statistic by definition counts plays that result in or nearly result in first downs as being successful. I have no doubt that such a statistic would be highly predictive of first downs. I would also have no doubt that past ability to gain first downs in and of itself is also highly predictive of future ability to gain first downs.

 

[TheSport78]

Correlation between YPA and winning percentage. You can run the analysis in excel.

It's statistically proven that running better is not useful. It's a knowable, indisputable fact. I'm not sure we need reasons as to why it would "seem" to be true given it's a provable certainty.

The success rate statistic by definition counts plays that result in or nearly result in first downs as being successful. I have no doubt that such a statistic would be highly predictive of first downs. I would also have no doubt that past ability to gain first downs in and of itself is also highly predictive of future ability to gain first downs.

Okay, let's look at specifically the Cowboys, since it's essentially pointless to make a generalized statements like "you don't need to run the ball in the NFL," or "passing well is the only thing that matters."

Cowboys rushing offensive rankings since Romo became starter in 06':

2006 = #13
2007 = #17
2008 = #21
2009 = #7
2010 = #16
2011 = #15
2012 = #31

Oh gee, look at that. The ONE year the Cowboys had a top 10 rushing attack, they happened to have won their ONLY playoff game in the past 17 years. Yup, running the football doesn't equivalate to success in the NFL.

And I'll add to that. Romo only threw NINE INTERCEPTIONS that year (2009).

Better execution with running the football = less turnover prone Romo = less pressure on defense to generate turnovers = more W's = playoff and hopefully Super Bowl wins

BALANCE is essential in the NFL, and that's all I really have to say.

 

[theogt]

Okay, let's look at specifically the Cowboys, since it's essentially pointless to make a generalized statements like "you don't need to run the ball in the NFL," or "passing well is the only thing that matters."

Cowboys rushing offensive rankings since Romo became starter in 06':

2006 = #13
2007 = #17
2008 = #21
2009 = #7
2010 = #16
2011 = #15
2012 = #31

Oh gee, look at that. The ONE year the Cowboys had a top 10 rushing attack, they happened to have won their ONLY playoff game in the past 17 years. Yup, running the football doesn't equivalate to success in the NFL.

And I'll add to that. Romo only threw NINE INTERCEPTIONS that year (2009).

Better execution with running the football = less turnover prone Romo = less pressure on defense to generate turnovers = more W's = playoff and hopefully Super Bowl wins

BALANCE is essential in the NFL, and that's all I really have to say.

This is anecdotal evidence, not statistical analysis.

 

[TheRomoSexual]

This is anecdotal evidence, not statistical analysis.

To be fair, you've only talked about statistics. You haven't presented any.

 

[theogt]

To be fair, you've only talked about statistics. You haven't presented any.

That's because the statistics I'm referencing are well known statistics. If you'd like me to run some, I can put it together in excel fairly quickly.

 

[theogt]

Some correlation stats:

http://www.advancednflstats.com/2007/07/what-makes-teams-win-3.html

 

[xwalker]

I work with a large number of people that are professional statistical analysts. Many of these people have a PhD in the field and many have decades of experience. If you spend time with these people, you would quickly understand that only the most simplistic of statistics are useful to non-professionals. Simplistic correlations like Yards per run play are basically worthless as a stand alone statistic. Using statistics to determine the causes for winning football games is a complex multi-variable equation that can't be accurately represented by simplistic single variables like Yards/Run.

I'll give one very simplistic example. Teams that are significantly ahead normally run the ball more in the 4th quarter. There is a good chance that the winning team in this scenario has a higher number of run attempts per game than the league average per game. It is also possible that they have a lower Yards/Attempt than average because they are just trying to run out the clock. They could even be using their 2nd and 3rd string RBs in this situation. The opposite could also be true based on various other variables. There are other situations where defenses play more or less "8 men in the box" against the run depending on the situation.

The point is that none of these other variables are accounted for in the simplistic Yards/Run-Attempt vs Winning correlations.

For professional statistical analysts, it all comes down to cause and effect. Is a low or high Yards/Attempt a cause of winning or does winning cause the low or high Yards/Attempt?

I know for a fact that some teams (probably all) have access to complex statistical models that go far beyond any data or equation that you will be able to access for free. I know people that work for companies that create these models and develop tools for processing the data.

It is obvious that coaches in the modern NFL believe that the passing game is more important than the running game in terms of winning; however, I believe that it is much closer to 60-40 than it is to 100-0 in terms of how they see the importance of the Pass-Run ratio.

 

[hra8700]

I work with a large number of people that are professional statistical analysts. Many of these people have a PhD in the field and many have decades of experience. If you spend time with these people, you would quickly understand that only the most simplistic of statistics are useful to non-professionals. Simplistic correlations like Yards per run play are basically worthless as a stand alone statistic. Using statistics to determine the causes for winning football games is a complex multi-variable equation that can't be accurately represented by simplistic single variables like Yards/Run.

I'll give one very simplistic example. Teams that are significantly ahead normally run the ball more in the 4th quarter. There is a good chance that the winning team in this scenario has a higher number of run attempts per game than the league average per game. It is also possible that they have a lower Yards/Attempt than average because they are just trying to run out the clock. They could even be using their 2nd and 3rd string RBs in this situation. The opposite could also be true based on various other variables. There are other situations where defenses play more or less "8 men in the box" against the run depending on the situation.

The point is that none of these other variables are accounted for in the simplistic Yards/Run-Attempt vs Winning correlations.

For professional statistical analysts, it all comes down to cause and effect. Is a low or high Yards/Attempt a cause of winning or does winning cause the low or high Yards/Attempt?

I know for a fact that some teams (probably all) have access to complex statistical models that go far beyond any data or equation that you will be able to access for free. I know people that work for companies that create these models and develop tools for processing the data.

It is obvious that coaches in the modern NFL believe that the passing game is more important than the running game in terms of winning; however, I believe that it is much closer to 60-40 than it is to 100-0 in terms of how they see the importance of the Pass-Run ratio.

Agree. The idea of saying something is "proven" from a simple univariate regression is kind of silly. Nothing is ever proven through analytics, all analytics do is test hypotheses. When something doesnt look right, then if youre lucky you can find a new variable that you never thought about. But youre right, nfl teams have a great deal of resources at their disposal, and if you watch demitroff talking with brian burke at sloan its clear they have everything he does and quite a bit more at their disposal. And teams still like to run the ball...

There are definitely things teams do that are wrong, but most things people criticize teams for its the teams that probably have it right more often than not.

 

[TheSport78]

Correlation does not always equal causation. Look back at the only time Dallas had a top ten running game. We won our only playoff game in the past 17 years and Romo only threw 9 picks.

 

[bayeslife]

Correlation does not always equal causation. Look back at the only time Dallas had a top ten running game. We won our only playoff game in the past 17 years and Romo only threw 9 picks.

Romo throwing only 9 picks is favoring the argument of passing the ball, not running. I don't even know a way to prove that a better running game somehow makes a quarterback throw better. It certainly is not the case in Washington, I'll tell you that much.

 

[TheRomoSexual]

Romo throwing only 9 picks is favoring the argument of passing the ball, not running. I don't even know a way to prove that a better running game somehow makes a quarterback throw better. It certainly is not the case in Washington, I'll tell you that much.

A strong run game can most certainly help the passing game. It allows the quarterback to ease into the game, enhances the effectiveness of the play-action, and prevents the defense from over-relying on certain blitz and coverage formations.

That's the problem with applying raw statistics to football -- they completely devalue the variable human component, especially when you are trying to draw conclusions about a singular based from statistics from teams across the NFL.

 

[SMCowboy]

I know enough that to be effective with it - Tampa used two anchoring (not fat but low center gravity guys who handle doubles) DT who could also penetrate. Offenses will run thru Dallas interior which also means more unnecessary punishment for Sean Lee as Guards will be getting to second level fairly easy.

The more I listen to reports the more I realize Dallas wasn't ready for the switch to 4-3. The draft could have helped things but Dallas didn't even make an attempt. They have become a reactionary team and only address things when it cost them games/seasons.

You have proven to have ZERO idea what you are talking about. Being able to Anchor mean NOTHING in the Tampa 2, when you teach your guys on every snap to attack a gap. The DL in the Tampa 2 have one responsibility, to get after the QB. And if you happen to run into the ball carrier along the way, that is great.

Infact, even if they do happened to get double teamed, and blown off the ball, that is fine. Because, in doing so they still did their job, which was to keep the OL off the LB's. The Tampa 2 is built on DT's rushing that passer, or LB's stopping the run.

 

[TheSport78]

Romo throwing only 9 picks is favoring the argument of passing the ball, not running. I don't even know a way to prove that a better running game somehow makes a quarterback throw better. It certainly is not the case in Washington, I'll tell you that much.

LOL you serious?

 

[CF74]

Romo throwing only 9 picks is favoring the argument of passing the ball, not running. I don't even know a way to prove that a better running game somehow makes a quarterback throw better. It certainly is not the case in Washington, I'll tell you that much.

 

[theogt]

I work with a large number of people that are professional statistical analysts. Many of these people have a PhD in the field and many have decades of experience. If you spend time with these people, you would quickly understand that only the most simplistic of statistics are useful to non-professionals. Simplistic correlations like Yards per run play are basically worthless as a stand alone statistic. Using statistics to determine the causes for winning football games is a complex multi-variable equation that can't be accurately represented by simplistic single variables like Yards/Run.

I'll give one very simplistic example. Teams that are significantly ahead normally run the ball more in the 4th quarter. There is a good chance that the winning team in this scenario has a higher number of run attempts per game than the league average per game. It is also possible that they have a lower Yards/Attempt than average because they are just trying to run out the clock. They could even be using their 2nd and 3rd string RBs in this situation. The opposite could also be true based on various other variables. There are other situations where defenses play more or less "8 men in the box" against the run depending on the situation.

The point is that none of these other variables are accounted for in the simplistic Yards/Run-Attempt vs Winning correlations.

For professional statistical analysts, it all comes down to cause and effect. Is a low or high Yards/Attempt a cause of winning or does winning cause the low or high Yards/Attempt?

I know for a fact that some teams (probably all) have access to complex statistical models that go far beyond any data or equation that you will be able to access for free. I know people that work for companies that create these models and develop tools for processing the data.

It is obvious that coaches in the modern NFL believe that the passing game is more important than the running game in terms of winning; however, I believe that it is much closer to 60-40 than it is to 100-0 in terms of how they see the importance of the Pass-Run ratio.

I believe the goal is prediction, in which case there isn't a more useful, simple tool than correlation. There's really no need to complicate it just to sound smart. It's not rocket science.

 

[hra8700]

I believe the goal is prediction, in which case there isn't a more useful, simple tool than correlation. There's really no need to complicate it just to sound smart. It's not rocket science.

"running yards per attempt is only weakly correlated with winning in the NFL. "

"having a good running game does NOT contribute winning in the NFL."

These two statements are not identical. You have to make dozens of leaps to get to that second statement. That is the nuance of statistical analysis. The first statement, which is true, is not actionable information, it's simply hypothesis generating.

 

[Slamman]

Some correlation stats:



[/QUOTE]

I enjoy stats and many stats are informative. Of course, correlations and cause and effect are not synonymous. IMO, when we break down NFL football with all ts parity down to its simplest form, I still think TO ratio is probably the best predictors of who wins. The teams that won the TO battle were a perfect 11-0 in the playoffs last year. The Cowboys were undefeated last year when they won the TO battle and 3-1 when they tied the opposing team in TOs. Dallas was 1-7 in games they lost the TO battle. With all the parity in the NFL, the team with the most opportunity to score is likely going to win.

The other thing I'd like to point out is that if we were going to predict right now which team will represent the NFC in the SB next year, the vast majority of predictions would come down to SF and Seattle. Yet, if we are purely looking at passing... I'd say GB and ATL has better passing attacks and certainly better offensive weapons. Even the Cowboys have arguably better offensive weapons. So, what sets SF and SEA apart? Their trenches on both sides of the ball combined with really good motivational coaching. I don't know if there's a stat that can truly measure that. But, we know it when we see it.

 

[theogt]

"running yards per attempt is only weakly correlated with winning in the NFL. "

"having a good running game does NOT contribute winning in the NFL."

These two statements are not identical. You have to make dozens of leaps to get to that second statement. That is the nuance of statistical analysis. The first statement, which is true, is not actionable information, it's simply hypothesis generating.

Annd yet it's so highly predictive. I'm okay with small leaps in logic as long as it turns out true time after time after time.

This is just a more advanced willful ignorance position for those that are stuck in the past.

 

[hra8700]

Annd yet it's so highly predictive. I'm okay with small leaps in logic as long as it turns out true time after time after time.

This is just a more advanced willful ignorance position for those that are stuck in the past.

Actually, you would say it's not VERY predictive, there is a slight positive correlation and Brian burke actually includes it in his model, just with a smaller coefficient than passing.

But let's say you're right and there's no correlation whatsoever and its not at all predictive of wins.

What do you propose? Making a claim without detailing its implications is vague and meaningless. If this were really some immutable 'fact' as you claim, that running does not lead to winning, then it should follow that a team that does not run the ball ever would win just as much, no?

I think its pretty clear that that is not the case. After all, variance of passing success is much higher than running, and in a system where you have 4 tries at a first down, you want consistent low variance yardage. Ad if teams new you never ran the ball, obviously the defensive fronts would change to prevent the pass.

 

[theogt]

Actually, you would say it's not VERY predictive

Correlation is itself a predictive measure in football, both in measuring what contributes to winning and what doesn't.

there is a slight positive correlation and Brian burke actually includes it in his model, just with a smaller coefficient than passing.

I believe you're referencing the "success rate" statistic and that's been addressed in this thread already.

But let's say you're right and there's no correlation whatsoever and its not at all predictive of wins.

What do you propose? Making a claim without detailing its implications is vague and meaningless. If this were really some immutable 'fact' as you claim, that running does not lead to winning, then it should follow that a team that does not run the ball ever would win just as much, no?

I think its pretty clear that that is not the case. After all, variance of passing success is much higher than running, and in a system where you have 4 tries at a first down, you want consistent low variance yardage. Ad if teams new you never ran the ball, obviously the defensive fronts would change to prevent the pass.

Yes, it's clear that is not the case. And that's been addressed in this thread.