The best way to win in this league

[Mikie2times]

I do a lot of football research and I pass some of the stuff I find interesting along. What I did this time was run a Pearson linear correlation test with several different NFL stats and compared the relationship they have to win%. For those a little fuzzy on correlation methods all you really need to know is the closer the correlation is to 1 the stronger the relationship between the two variables is. The closer the correlation is to -1 the stronger the inverse relationship of the varibles are. Here are some of the most obvious correlations I found based on the 2005 and 2004 seasons.

 

Offense

Rush Attempts per Game .6305 -Playoff teams in the top 10) 2005:6....2004:5

Rushing Yards Per Game .6308- Playoff teams in the top 10) 2005:7....2004:8

Passing Yards Per Attempt .5859- Playoff teams in the top 10) 2005:7....2004:8

 

Conclusions- Rushing Yards Per Attempt had a moderate correlation at .4054, while passing yards per game had practically no correlation at .1871. If you pass for a lot of yards it most likely means you’re playing from behind or lack balance. Either way this shows no evidence that lots of passing yards will win you games. What it does show is if you pass with efficiency, and a higher yard average per attempt the passing game will matter on offense. In 2005 Buffalo ranked 20th in rushing attempts, 20th in yards rushing per game, and 27th in yards per pass attempt.

 

Defense

Rush Attempts Per Game -.7077- Playoff teams in the top 10) 2005:6....2004:6

Rushing Yards Per Game -.6372- Playoff teams in the top 10) 2005:6....2004:6

 

Conclusions- The importance of Passing Yards Per Attempt dropped off substantially on the defensive side of the ball. On offense it nearly had a .60 correlation, on defense only a -.30. Like offense, passing yardage also didn’t show a strong correlation with win% at -.1143. In 2005 Buffalo ranked 25th in Rush Attempts per Game on defense, and 31st in rushing yards per game allowed.

 

This evidence suggests stopping the run is the most important aspect of football. Second to that is being able to run the football yourself, and finally being able to pass efficiently also plays a strong role. Now obviously football is not as cut and dry as that, but what this does show is that making the playoffs and being successful is most attainable by being good in these areas.

[IDBillzFan]

It's stunning that you would take the time to do that kind of research. Kudos for the effort, dude.

 

You must've been pissed when you went through all that and found it that to win you need to run the ball well and keep the other team from running the ball well.

[Mikie2times]

It's stunning that you would take the time to do that kind of research. Kudos for the effort, dude.

 

You must've been pissed when you went through all that and found it that to win you need to run the ball well and keep the other team from running the ball well.

710270[/snapback]

I got a pretty healthy database built up which allows be to confirm the obvious. I also hear smoking is bad for you.

[PatPatPatSack]

It's stunning that you would take the time to do that kind of research. Kudos for the effort, dude.

 

You must've been pissed when you went through all that and found it that to win you need to run the ball well and keep the other team from running the ball well.

710270[/snapback]

OTOH, maybe the correlation is there ONLY because that is the typical/usual game plan. And the correlation is measuring the success of the game plan vs wins. So that simply said, if it's what you set out to do, and you are successful, you will probably win. What I am suggesting is, what happens when the gameplan is not run/stop the run?

 

What about the west coast which I understand to be more of a passing completion offense. In the case of a pure west coast, what would the measurement of effectiveness of the game plan be?

 

I often have pondered this. Let say team a starts the season in typical fashion. We are going to run/stop the run. And the fans all chime in, yeah baby. But after 2 games it becomes apparent that the team cannot do this. Should team a have a new more suitable strategy geared toward something they could do. Or should they continue to follow the mantra. R/STR. If you could identify that type of team and evaluate their success at the new game plan, then you could evaluate the win/loss thing too.

 

Then agin maybe I ponder too much.

[BuffaloRebound]

I think you may have a cause-and-effect problem with that research. Teams that are winning naturally run more and teams that are behind get run on. I would be more interested to see the relationship between 3rd down conversion rates on both sides of the ball and winning %.

[bbfan54]

Yeah I believe that has more to do with teams running out the clock at the end of the game.

[RLflutie7]

To further illustrate the facts: Top five rush defense teams in 2005:

1. San Diego;

2. Denver;

3. Pittsburgh;

4. Carolina;

5. Seattle.

 

Four of those five teams were in the NFC and AFC championship games.

[Tortured Soul]

I think you may have a cause-and-effect problem with that research.  Teams that are winning naturally run more and teams that are behind get run on.  I would be more interested to see the relationship between 3rd down conversion rates on both sides of the ball and winning %.

710275[/snapback]

All he's showing is correlation, not causation.

[Mikie2times]

I think you may have a cause-and-effect problem with that research.  Teams that are winning naturally run more and teams that are behind get run on.  I would be more interested to see the relationship between 3rd down conversion rates on both sides of the ball and winning %.

710275[/snapback]

 

Good post. I agree that this factor does influence the data, but I don't feel as if it explains the data.

 

If you wanted to remove these factors in the rushing stats you could by using rushing yards per attempt for offense, but that would inevitably lead to more problems with the data. For example Philadelphia last year was tied for 16th in rushing yards per attempt, but that's only because they had teams playing the pass so often. In actual rushing yardage they ranked 28th near the bottom of the league.

 

It's clear that rushing success on offense is not only about averages but repetition. Repetition wears down defenses, and it allows the offense a chance to move the ball with little chance of a turnover. A team like Philadelphia runs the ball as a way to keep teams from dropping everybody, but when they need to run they don't do a very good job at it. Last years top 10 rushing teams included Atlanta, Denver, Seattle, Kansas City, Pittsburgh, NY Giants, Washington, Chicago, San Diego and Jacksonville. Just about everybody on that list is known for running the ball, and they do it well. The vast majority of these teams win because they run; they don't run because they're winning.

 

The passing yardage correlation is heavily influenced by the type of behavior you speak of thus it has no correlation. But passing yards per attempt is not influenced by much, if anything. So it should be at least somewhat surprising that correlation difference of this stat varies so much from offense to defense.

 

As for your 3rd down stat request hear it is based on 04 and 05

Offense .4761

Defense -.1021

[silvermike]

If it's broken down in your database, what about first-half rushing yards and attempts? That should sift out plenty of the coming back/running out the clock parts of it.

[Orton's Arm]

What I did this time was run a Pearson linear correlation test

Not the giggity giggity goo I expected from that avatar!

 

As was pointed out, one thing your data are picking up is that teams with the lead tend to run more. So the number of rushes per game has to be ignored for this reason. What I found interesting is that passing yards per attempt had a higher correlation than rushing yards per attempt. This may not mean much though, as the sample size is probably too small to allow conclusions at 95% or even 90% confidence anyway.

[kenny3000]

can you check the correlation of flip-flopping (if you will) the starting quarterback to making it to an AFC/NFC championship?

[Billsguy]

What I did this time was run a Pearson linear correlation test with several different NFL stats and compared the relationship they have to win%.

 

Conclusions- This evidence suggests stopping the run is the most important aspect of football.

710264[/snapback]

 

 

As you know, correlation does not imply causality. Therefore, the fact that a certain factor is highly correlated to winning does not mean that the factor in question is the one that caused the victory.

 

For instance, a team with a great rush defense statistically might also have a great offense. Since their offense is great, they score often and build up leads against their opponents.

 

The opponents are therefore forced to abandon the run game and throw more often in an effort to catch up on the scoreboard.

 

The end result is that the high scoring team wins many games, but it is not because they have a great rush defense. (in this case, nobody has the luxury to run against them because they are behind in the score)

 

A conclusion that says "stopping the run is the most important aspect" in this case would have a high statistical correlation and yet be completely false!!!!!

 

Your statistical analysis may be interesting to some people, but it is simply wrong to reach your conclusions based on linear regression. By wrong I mean invalid and inaccurate.

 

Stopping the run is certainly important, but you can't conclude that using the technique you used. Sorry, but true.

[Bill from NYC]

I do a lot of football research and I pass some of the stuff I find interesting along. What I did this time was run a Pearson linear correlation test with several different NFL stats and compared the relationship they have to win%. For those a little fuzzy on correlation methods all you really need to know is the closer the correlation is to 1 the stronger the relationship between the two variables is. The closer the correlation is to -1 the stronger the inverse relationship of the varibles are.  Here are some of the most obvious correlations I found based on the 2005 and 2004 seasons.

 

Offense

Rush Attempts per Game  .6305 -Playoff teams in the top 10) 2005:6....2004:5

Rushing Yards Per Game    .6308- Playoff teams in the top 10) 2005:7....2004:8

Passing Yards Per Attempt .5859- Playoff teams in the top 10) 2005:7....2004:8

 

Conclusions- Rushing Yards Per Attempt had a moderate correlation at .4054, while passing yards per game had practically no correlation at .1871. If you pass for a lot of yards it most likely means you’re playing from behind or lack balance. Either way this shows no evidence that lots of passing yards will win you games.  What it does show is if you pass with efficiency, and a higher yard average per attempt the passing game will matter on offense.  In 2005 Buffalo ranked 20th in rushing attempts, 20th in yards rushing per game, and 27th in yards per pass attempt.

 

Defense

Rush Attempts Per Game  -.7077- Playoff teams in the top 10) 2005:6....2004:6

Rushing Yards Per Game -.6372- Playoff teams in the top 10) 2005:6....2004:6

 

Conclusions- The importance of Passing Yards Per Attempt dropped off substantially on the defensive side of the ball. On offense it nearly had a .60 correlation, on defense only a -.30. Like offense, passing yardage also didn’t show a strong correlation with win% at -.1143. In 2005 Buffalo ranked 25th in Rush Attempts per Game on defense, and 31st in rushing yards per game allowed.

 

This evidence suggests stopping the run is the most important aspect of football. Second to that is being able to run the football yourself, and finally being able to pass efficiently also plays a strong role. Now obviously football is not as cut and dry as that, but what this does show is that making the playoffs and being successful is most attainable by being good in these areas.

710264[/snapback]

 

Very good post. Here is another theory that I subscribe to, and I think that it ties into yours.

 

Please check out the success rate of teams who select 1st round guards.

[Mikie2times]

Not the giggity giggity goo I expected from that avatar! 

 

As was pointed out, one thing your data are picking up is that teams with the lead tend to run more.  So the number of rushes per game has to be ignored for this reason.  What I found interesting is that passing yards per attempt had a higher correlation than rushing yards per attempt.  This may not mean much though, as the sample size is probably too small to allow conclusions at 95% or even 90% confidence anyway.

710344[/snapback]

Again I understand it’s picking up instances of teams running out the clock but let's please keep this in perspective.

 

The league wide average for rush % over this span is 46.6%, and in the last two years (playoffs and regular season combined) running over 50% of the time was accomplished only 19 times. Of that group only 6 teams averaged less then 4 yards a carry, and only last years Panthers had a respectful Yards Per Pass Attempt average. To me that indicates that teams either run a lot because they do it well, or they have to run because they have such an inefficient passing game.

 

As for your other comments passing yards per attempt is more of a double statistic. This type of accuracy makes it a perfect compliment to this type of testing. Think of it like TD/INT ratio or SO/BB in baseball. Naturally if you are accurate the yards per attempt will increase but in order for it to increase beyond the norm you still need to complete those passes for decent yardage. The worst players in this category end up always being rookies, while the best almost always have solid TD/INT ratios.

 

With rushing yards per attempt a passing team can excel because they run on occasion to set up the pass. Teams like Philadelphia and Minnesota have historically been strong in rushing yards per attempt, but weak in overall rushing attempts and rushing yards. Since RYPA doesn't accurately measure a team’s true rushing ability it will naturally lose correlation.

 

Also my database goes back to 1997 and I have run these tests many times. Take it for what it's worth but the correlations come up with similar results in the categories I mentioned.

[Mikie2times]

As you know, correlation does not imply causality. Therefore, the fact that a certain factor is highly correlated to winning does not mean that the factor in question is the one that caused the victory.

 

For instance, a team with a great rush defense statistically might also have a great offense. Since their offense is great, they score often and build up leads against their opponents.

 

The opponents are therefore forced to abandon the run game and throw more often in an effort to catch up on the scoreboard.

 

The end result is that the high scoring team wins many games, but it is not because they have a great rush defense. (in this case, nobody has the luxury to run against them because they are behind in the score)

 

A conclusion that says "stopping the run is the most important aspect" in this case would have a high statistical correlation and yet be completely false!!!!!

 

Your statistical analysis may be interesting to some people, but it is simply wrong to reach your conclusions based on linear regression. By wrong I mean invalid and inaccurate.

 

Stopping the run is certainly important, but you can't conclude that using the technique you used. Sorry, but true.

710349[/snapback]

Your example is valid, but in this sampling only a couple of the teams would produce this result. Last years Colts team would be a perfect example of this. They built leads and teams wouldn't run so they looked like they had a decent run defense but the reality was different. These teams escape this type of testing but they're also rare examples. The vast majority of the league would not fit this profile, and the vast majority of the league's games are competitive.

[Dawgg]

This is precisely why the Bills won a lot of games under the Wade Phillips era. He was the first (and last) Bills coach to stress first and foremost stopping the run on the defensive side of the ball and for the most part, that's what the Bills did. During those playoff years, they were among the NFL elete in rushing and stopping the rush.

 

 

I do a lot of football research and I pass some of the stuff I find interesting along. What I did this time was run a Pearson linear correlation test with several different NFL stats and compared the relationship they have to win%. For those a little fuzzy on correlation methods all you really need to know is the closer the correlation is to 1 the stronger the relationship between the two variables is. The closer the correlation is to -1 the stronger the inverse relationship of the varibles are.  Here are some of the most obvious correlations I found based on the 2005 and 2004 seasons.

 

Offense

Rush Attempts per Game  .6305 -Playoff teams in the top 10) 2005:6....2004:5

Rushing Yards Per Game    .6308- Playoff teams in the top 10) 2005:7....2004:8

Passing Yards Per Attempt .5859- Playoff teams in the top 10) 2005:7....2004:8

 

Conclusions- Rushing Yards Per Attempt had a moderate correlation at .4054, while passing yards per game had practically no correlation at .1871. If you pass for a lot of yards it most likely means you’re playing from behind or lack balance. Either way this shows no evidence that lots of passing yards will win you games.  What it does show is if you pass with efficiency, and a higher yard average per attempt the passing game will matter on offense.  In 2005 Buffalo ranked 20th in rushing attempts, 20th in yards rushing per game, and 27th in yards per pass attempt.

 

Defense

Rush Attempts Per Game  -.7077- Playoff teams in the top 10) 2005:6....2004:6

Rushing Yards Per Game -.6372- Playoff teams in the top 10) 2005:6....2004:6

 

Conclusions- The importance of Passing Yards Per Attempt dropped off substantially on the defensive side of the ball. On offense it nearly had a .60 correlation, on defense only a -.30. Like offense, passing yardage also didn’t show a strong correlation with win% at -.1143. In 2005 Buffalo ranked 25th in Rush Attempts per Game on defense, and 31st in rushing yards per game allowed.

 

This evidence suggests stopping the run is the most important aspect of football. Second to that is being able to run the football yourself, and finally being able to pass efficiently also plays a strong role. Now obviously football is not as cut and dry as that, but what this does show is that making the playoffs and being successful is most attainable by being good in these areas.

710264[/snapback]

[Dibs]

Please check out the success rate of teams who select 1st round guards.

710351[/snapback]

From 1982-2003 there have been 31 players drafted as guards in the 1st round.

Of those 31, 9 have gone on to play in a Superbowl...or 29%

On the surface, that might sound O.K. but.....

Lets assume the average playing age of an NFL 1st Rounder is 5 years.

In any given 5 year span, any guards chance of playing in a SB is 5 in 16...or 31%

As you can see, on straight averages the 1st round guards are below the average in SB appearances(& most were playing in pre-expansion times making it worse).

 

My interpretation of these stats are....

Drafting a guard in the first round does not make any difference on weather the team will make the SB or not. I think it is just that lately(Steelers & Seahawks) it seems to work....over the last 20 years however, it simply equals the average at best.

 

I can list all the picks & the ones that played in the SB if people want. I chose not to to save on scrolling.

Edited June 19, 2006 by Dibs

[Albany,n.y.]

can you check the correlation of flip-flopping (if you will) the starting quarterback to making it to an AFC/NFC championship?

710345[/snapback]

It worked in the AFC Championship game in Miami in 1993. Will you please stop this "Frank Reich should have started" nonsense. Kelly won the game, give it up!

[Bill from NYC]

From 1982-2003 there have been 31 players drafted as guards in the 1st round.

Of those 31, 9 have gone on to play in a Superbowl...or 29%

On the surface, that might sound O.K. but.....

Lets assume the average playing age of an NFL 1st Rounder is 5 years.

In any given 5 year span, any guards chance of playing in a SB is 5 in 16...or 31%

As you can see, on straight averages the 1st round guards are below the average in SB appearances(& most were playing in pre-expansion times making it worse).

 

My interpretation of these stats are....

Drafting a guard in the first round does not make any difference on weather the team will make the SB or not.  I think it is just that lately(Steelers & Seahawks) it seems to work....over the last 20 years however, it simply equals the average at best.

 

I can list all the picks & the ones that played in the SB if people want.  I chose not to to save on scrolling.

710372[/snapback]

 

Good points. The Steelers and Seahawks were a fluke.