The Cowboys problem is not that Complicated

[Outlaw Heroes]

Originally Posted by AdamJT13

Run success does not "correlate fairly strongly." It's still a low to moderate-at-best correlation -- and much lower than pass success.

Lower that pass success and lower than pass efficiency, but as I say, I don't think that's really in dispute.

It's also fair to say that R values of around .4 are "moderate-at-best", but then I think most statisticians would say that R values don't really exceed "moderate" until you get up over .7. By that standard, pass success and pass efficiency also yield moderate corrrelations.

Originally Posted by AdamJT13

His point wasn't that run success correlates fairly strongly, it's that run success correlates more than simple YPC, which is true.

Certainly he's saying that run success correlates with wins more strongly than YPC. But I think you're understating what he's saying. From the article:

The bottom line is that we should pay attention to run SR. It’s very likely a predictor [emphasis added] of passing performance and of overall team success.

Originally Posted by AdamJT13

I'm guessing that a good portion of that difference could be attributed to the ability to convert on short-yardage runs.

You're surely right about this. I trust you don't see that as a limitation in run SR? I personally think it's a strength. Any model that tries to predict team success on the basis of effective running without placing significant emphasis on the ability to convert in short-yardage situations (a failing of YPC) is ipso facto of questionable value, in my view.

[MichaelWinicki]

Originally Posted by AdamJT13

Defensive teams are fooling themselves (and hurting themselves) if they think they "must stop the run." The Dolphins did just that and got blown out because they couldn't pass effectively or stop the pass.

And as far as the running game helping the passing game, two of the Texans' three best running games also happen to be two of Schaub's three worst passing games. A theory is just a theory until it is proved on the field.

I think it's going to take fans some time to wrap their heads around the emerging world of football statistics.

I remember getting my first "Bill James Abstract" back in 1981 and the result changed my view on how baseball, and the stats it was using at the time to interpret the game.

His research turned the game on its head so to speak.

Up until that time "we" were told that guys like Mickey Rivers (high average, no walks, lots of speed) were the "ideal" leadoff men... basically neglecting on-base percentage.

We also heard things like "Reggie Jackson performs better in front of bigger crowds".

So many old baseball-clichés were sent to the trash-bin because of Bill James.


30 years later, we are seeing the same in football.

Many of us have heard the tired, old clichés for literally decades...

"To win you need to be able to run the ball and stop the run"

"The game is won in the trenches"

And stuff like that.


Now we are finding out that the old axiom's weren't necessarily true and it's shaking the foundation of what many of us thought was "winning football".

[AdamJT13]

Originally Posted by Outlaw Heroes

Lower that pass success and lower than pass efficiency, but as I say, I don't think that's really in dispute.

It's also fair to say that R values of around .4 are "moderate-at-best", but then I think most statisticians would say that R values don't really exceed "moderate" until you get up over .7. By that standard, pass success and pass efficiency also yield moderate corrrelations.

The correlations that I've seen for run success and similar metrics are usually 0.30 to 0.40. For pass success and similar metrics, they're 0.65 to 0.85. So the high end of the rushing correlations would be the low end of a "moderate" correlation, and the low end of the passing correlations would be the high end of a "moderate" correlation. But the median for rushing correlations would be considered more a "low" correlation, and the median for passing would be considered more a "high" correlation. At any rate, the difference is quite large.

Of course, most of these correlations are based on only one side of the ball at a time and are based on season-long averages and win totals. I'm more interested in the numbers for both sides of the ball and their correlation with winning individual games.

For example, teams that have passed more effectively in a game are 94-24 (.797) this season. Teams that have rushed more effectively (YPC) are 64-54 (.542).

Teams that have passed more effectively AND rushed more effectively are 50-10 (.833). Teams that have passed more effectively but rushed LESS effectively are 44-14 (.759).

Looking at those numbers conversely, that means that teams that have rushed more effectively but passed LESS effectively are 14-44 (.241). And teams that have rushed AND passed LESS effectively are 10-50 (.167).

Those winning percentages are about the same as they are every season -- about 80 percent for passing and a little more than 50-50 for rushing. Given that teams win about 17 percent of the time even if they're statistically worse at both rushing and passing in that game (and therefore lose about 17 percent of the time even if they're statistically better at both during that game), the numbers show that being able to pass better is overwhelmingly more important than being able to run better.

Also, note that the teams that ran the ball better passed it better in 60 games and passed it worse in 58 games -- again barely more than a 50-50 split. Running the ball better than your opponent has very little effect on being able to pass it better or winning.

[goshan]

Originally Posted by AdamJT13

The correlations that I've seen for run success and similar metrics are usually 0.30 to 0.40. For pass success and similar metrics, they're 0.65 to 0.85. So the high end of the rushing correlations would be the low end of a "moderate" correlation, and the low end of the passing correlations would be the high end of a "moderate" correlation. But the median for rushing correlations would be considered more a "low" correlation, and the median for passing would be considered more a "high" correlation. At any rate, the difference is quite large.

Of course, most of these correlations are based on only one side of the ball at a time and are based on season-long averages and win totals. I'm more interested in the numbers for both sides of the ball and their correlation with winning individual games.

[View Full Quote] For example, teams that have passed more effectively in a game are 94-24 (.797) this season. Teams that have rushed more effectively (YPC) are 64-54 (.542).

Teams that have passed more effectively AND rushed more effectively are 50-10 (.833). Teams that have passed more effectively but rushed LESS effectively are 44-14 (.759).

Looking at those numbers conversely, that means that teams that have rushed more effectively but passed LESS effectively are 14-44 (.241). And teams that have rushed AND passed LESS effectively are 10-50 (.167).

Those winning percentages are about the same as they are every season -- about 80 percent for passing and a little more than 50-50 for rushing. Given that teams win about 17 percent of the time even if they're statistically worse at both rushing and passing in that game (and therefore lose about 17 percent of the time even if they're statistically better at both during that game), the numbers show that being able to pass better is overwhelmingly more important than being able to run better.

Also, note that the teams that ran the ball better passed it better in 60 games and passed it worse in 58 games -- again barely more than a 50-50 split. Running the ball better than your opponent has very little effect on being able to pass it better or winning.

Powerful analysis of what matters.

[xwalker]

Originally Posted by AdamJT13

It is true. How well you pass the ball and stop the pass has *much* more effect on winning than how well you run the ball or stop the run.

You're using an Isolated Single-Independent-Variable Statistical model to determine winning.

Not only do you require multiple other variables to have any shot at an accurate model for winning, but there are multiple ways of defining "how well you pass" and "how well you run".

The Cowboy are 5th in Passing Yards per Game. Are they the 5th best team?

The Broncos have the best Pass per Attempt Average. Are they the best team?

[goshan]

Originally Posted by xwalker

You're using an Isolated Single-Independent-Variable Statistical model to determine winning.

Not only do you require multiple other variables to have any shot at an accurate model for winning, but there are multiple ways of defining "how well you pass" and "how well you run".

The Cowboy are 5th in Passing Yards per Game. Are they the 5th best team?

The Broncos have the best Pass per Attempt Average. Are they the best team?

Passing effectiveness does not equal passing yardage. You crack me up with your intro trying to sound like a statistical dude...then you follow it up with that silliness

Passing effectiveness the way Adam defines it is an aggregate metric, not a single-independent variable.

[Outlaw Heroes]

Originally Posted by AdamJT13

The correlations that I've seen for run success and similar metrics are usually 0.30 to 0.40. For pass success and similar metrics, they're 0.65 to 0.85. So the high end of the rushing correlations would be the low end of a "moderate" correlation, and the low end of the passing correlations would be the high end of a "moderate" correlation. But the median for rushing correlations would be considered more a "low" correlation, and the median for passing would be considered more a "high" correlation. At any rate, the difference is quite large.

Of course, most of these correlations are based on only one side of the ball at a time and are based on season-long averages and win totals. I'm more interested in the numbers for both sides of the ball and their correlation with winning individual games.

[View Full Quote] For example, teams that have passed more effectively in a game are 94-24 (.797) this season. Teams that have rushed more effectively (YPC) are 64-54 (.542).

Teams that have passed more effectively AND rushed more effectively are 50-10 (.833). Teams that have passed more effectively but rushed LESS effectively are 44-14 (.759).

Looking at those numbers conversely, that means that teams that have rushed more effectively but passed LESS effectively are 14-44 (.241). And teams that have rushed AND passed LESS effectively are 10-50 (.167).

Those winning percentages are about the same as they are every season -- about 80 percent for passing and a little more than 50-50 for rushing. Given that teams win about 17 percent of the time even if they're statistically worse at both rushing and passing in that game (and therefore lose about 17 percent of the time even if they're statistically better at both during that game), the numbers show that being able to pass better is overwhelmingly more important than being able to run better.

Also, note that the teams that ran the ball better passed it better in 60 games and passed it worse in 58 games -- again barely more than a 50-50 split. Running the ball better than your opponent has very little effect on being able to pass it better or winning.

Great stuff. Thanks for sharing.

[MichaelWinicki]

Originally Posted by goshan

Passing effectiveness does not equal passing yardage. You crack me up with your intro trying to sound like a statistical dude...then you follow it up with that silliness

Passing effectiveness the way Adam defines it is an aggregate metric, not a single-independent variable.

Yeah.

The passing yardage is fine.

But when you lead the NFL in picks...

Well, that's not too effective or efficient.

[AdamJT13]

Originally Posted by AdamJT13

For example, teams that have passed more effectively in a game are 94-24 (.797) this season. Teams that have rushed more effectively (YPC) are 64-54 (.542).

Teams that have passed more effectively AND rushed more effectively are 50-10 (.833). Teams that have passed more effectively but rushed LESS effectively are 44-14 (.759).

Looking at those numbers conversely, that means that teams that have rushed more effectively but passed LESS effectively are 14-44 (.241). And teams that have rushed AND passed LESS effectively are 10-50 (.167).

In Week 9, teams that passed more effectively went 14-0, and teams that rushed more effectively (YPC) went 7-7.

Here are the updated numbers --

Teams that have passed more effectively AND rushed more effectively are 57-10 (.851).

Teams that have passed more effectively but rushed LESS effectively are 51-14 (.785).

Teams that have rushed more effectively but passed LESS effectively are 14-51 (.215).

And teams that have rushed AND passed LESS effectively are 10-57 (.149).