Looked at NFL season-to-date winning percentages as a function of the following ratios:
Rushing Ratio Stat
Team's Offensive Rushing Yards per Carry / Team's Defensive Rushing Yards Allowed per Carry
Example: Arizona Cardinals 4.34 / 4.58 = .95 Rushing Ratio
Passing Ratio Stat
Team's Offensive Passing Yards per Attempt / Team's Defensive Passing Yards Allowed per Attempt
Example: Kansas City Chiefs 8.05 / 6.40 = 1.26 Passing Ratio
A regression analysis was run on 32 data pairs -- that is, a team's winning percentage as a function of each ratio above. Much to my surprise, the Rushing Ratio has little -- if any -- correlation to winning percentage. The r-squared value = 3.2%. In other words, 3.2% of a team's winning percentage is a function of its Running Ratio. This correlation is so low as to be almost meaningless. The passing ratio is a very different story -- the correlation is significant with an r-squared value = 30.9%. In other words, almost one-third of a team's winning percentage is a function of its Passing Ratio.
The most serious drawback to the above analysis is that there is a synergy between running and passing (both sides of the ball) and the correlations above do not account for that synergy. Moreover, this is just a snapshot of one season (through Week 12) and I'm too lazy right now to replicate this finding over past seasons. But for the time being, I am intrigued by the current findings.
************************************************************************************************************************************************************************************
Running Ratio:
Team/Rushing Yards per Carry/Rushing Yards per Carry Allowed/Ratio/Winning Percentage
Arizona/ 4.34/ 4.58/ 0.95/ 0.333
Atlanta/ 4.92/ 4.40/ 1.12/ 0.417
Baltimore/ 5.28/ 3.90/ 1.35/ 0.636
Buffalo/ 5.30/ 4.24/ 1.25/ 0.727
Carolina/ 4.45/ 4.59/ 0.97/ 0.333
Chicago/ 5.36/ 4.68/ 1.14/ 0.250
Cincinnati/ 3.86/ 4.27/ 0.91/ 0.636
Cleveland/ 4.85/ 4.84/ 1.00/ 0.364
Dallas/ 4.57/ 4.70/ 0.97/ 0.727
Denver/ 4.05/ 4.57/ 0.89/ 0.273
Detroit/ 4.59/ 5.22/ 0.88/ 0.364
Green Bay/ 4.80/ 4.97/ 0.97/ 0.333
Houston/ 4.11/ 4.91/ 0.84/ 0.136
Indianapolis/ 4.10/ 3.93/ 1.04/ 0.375
Jacksonville/ 4.86/ 4.24/ 1.15/ 0.364
Kansas City/ 4.68/ 4.35/ 1.08/ 0.818
LA Chargers/ 3.78/ 5.44/ 0.69/ 0.545
LA Rams/ 3.52/ 3.92/ 0.90/ 0.273
Las Vegas/ 5.16/ 4.28/ 1.20/ 0.364
Miami/ 4.00/ 4.46/ 0.90/ 0.727
Minnesota/ 4.25/ 4.35/ 0.98/ 0.818
New England/ 4.02/ 4.25/ 0.95/ 0.545
New Orleans/ 4.53/ 4.52/ 1.00/ 0.333
NY Giants/ 4.66/ 5.16/ 0.90/ 0.636
NY Jets/ 4.50/ 4.05/ 1.11/ 0.636
Philadelphia/ 4.73/ 4.68/ 1.01/ 0.909
Pittsburgh/ 4.27/ 3.99/ 1.07/ 0.364
San Francisco/ 4.41/ 3.32/ 1.33/ 0.636
Seattle/ 4.84/ 4.90/ 0.99/ 0.545
Tampa Bay/ 3.25/ 4.64/ 0.70/ 0.455
Tennessee/ 4.16/ 3.86/ 1.08/ 0.636
Washington/ 3.98/ 4.37/ 0.91/ 0.583
Passing Ratio:
Team/Passing Yards per Attempt/Defensive Passing Yards Allowed per Attempt/Ratio/Winning Percentage
Arizona/ 6.09/ 6.72/ 0.91/ 0.333
Atlanta/ 7.43/ 7.24/ 1.03/ 0.417
Baltimore/ 6.92/ 6.80/ 1.02/ 0.636
Buffalo/ 7.65/ 6.26/ 1.22/ 0.727
Carolina/ 6.75/ 6.21/ 1.09/ 0.333
Chicago/ 7.18/ 7.61/ 0.94/ 0.250
Cincinnati/ 7.77/ 6.45/ 1.20/ 0.636
Cleveland/ 7.08/ 6.88/ 1.03/ 0.364
Dallas/ 7.12/ 5.64/ 1.26/ 0.727
Denver/ 6.80/ 5.40/ 1.26/ 0.273
Detroit/ 7.36/ 7.53/ 0.98/ 0.364
Green Bay/ 6.99/ 6.66/ 1.05/ 0.333
Houston/ 6.40/ 7.14/ 0.90/ 0.136
Indianapolis/ 6.65/ 6.40/ 1.04/ 0.375
Jacksonville/ 6.88/ 7.07/ 0.97/ 0.364
Kansas City/ 8.05/ 6.40/ 1.26/ 0.818
LA Chargers/ 6.46/ 6.86/ 0.94/ 0.545
LA Rams/ 6.59/ 6.73/ 0.98/ 0.273
Las Vegas/ 7.05/ 7.38/ 0.96/ 0.364
Miami/ 8.49/ 6.62/ 1.28/ 0.727
Minnesota/ 6.73/ 7.78/ 0.87/ 0.818
New England/ 7.78/ 6.05/ 1.29/ 0.545
New Orleans/ 7.50/ 6.26/ 1.20/ 0.333
NY Giants/ 6.82/ 6.66/ 1.02/ 0.636
NY Jets/ 6.70/ 5.95/ 1.13/ 0.636
Philadelphia/ 8.07/ 5.35/ 1.51/ 0.909
Pittsburgh/ 6.08/ 7.28/ 0.84/ 0.364
San Francisco/ 7.75/ 6.09/ 1.27/ 0.636
Seattle/ 7.91/ 6.97/ 1.13/ 0.545
Tampa Bay/ 6.48/ 5.77/ 1.12/ 0.455
Tennessee/ 7.71/ 6.65/ 1.16/ 0.636
Washington/ 6.59/ 6.48/ 1.02/ 0.583
I think people look at this wrong. It isn't about passing vs. rushing in terms of wins and losses. Both are needed. It isn't one or the other. The way it should be looked at is that points come out of the passing game and being able to control a game comes from rushing the ball. The Gianta run the ball because they know they aren't going to win in a shootout. They are running to control the game and keep it close. The Chiefs pass the shit out of the ball because they can and will take their chances in a shootout.
You can't look at running in isolation (nor passing for that matter.) That's why I stated the following caveat:
"The most serious drawback to the above analysis is that there is a synergy between running and passing (both sides of the ball) and the correlations above do not account for that synergy."
Quote:
What would be the point of running the ball if it contributes such a low percentage to winning?
You can't look at running in isolation (nor passing for that matter.) That's why I stated the following caveat:
"The most serious drawback to the above analysis is that there is a synergy between running and passing (both sides of the ball) and the correlations above do not account for that synergy."
Sorry, I missed that. I have seen multiple fans use this data as evidence that running the ball is not an important part of winning which I do not agree with. Your caveat is spot on. There are grey areas that numbers just can't represent.
I also agree with Robbie's point. It matters how a team's roster is constructed; what play style fits their roster best? At one point in the league, running was all teams did until coaches realized how important passing is. Since then teams have always preferred to get that great qb to be able to rely less on the run game which makes it tougher for a defense to pin you down.
I think people look at this wrong. It isn't about passing vs. rushing in terms of wins and losses. Both are needed. It isn't one or the other. The way it should be looked at is that points come out of the passing game and being able to control a game comes from rushing the ball. The Gianta run the ball because they know they aren't going to win in a shootout. They are running to control the game and keep it close. The Chiefs pass the shit out of the ball because they can and will take their chances in a shootout.
You are absolutely correct that it is not about running vs passing. That is why I pointed out that a flaw in this analysis is that it does not account for the synergy between the two. The way to get a handle on this would be to run a multi-variable regression analysis with both Run and Pass Ratios as inputs. Once you get the mutli-variable regression equation, you could compare the coefficients for each ratio to see which one is higher.
Too funny! (But I did spend my career in media-marketing stat research.)
I think people look at this wrong. It isn't about passing vs. rushing in terms of wins and losses. Both are needed. It isn't one or the other. The way it should be looked at is that points come out of the passing game and being able to control a game comes from rushing the ball. The Gianta run the ball because they know they aren't going to win in a shootout. They are running to control the game and keep it close. The Chiefs pass the shit out of the ball because they can and will take their chances in a shootout.
+1.
In regard to your statement: "I also agree with Robbie's point. It matters how a team's roster is constructed; what play style fits their roster best?"
That is very true, but the impact on the correlation stat by any one team is minimal. And that's because the correlation in looking at the relationship between the Run (or Pass) Ratio on winning percentages across ALL 32 NFL teams -- not just one team in isolation.
That makes practical sense. If you can accrue more yards and points on fewer attempts, you have the ball less.
Which then might lead to more rushes to protect the lead.
Figure out what your opponent is not good at and attack it.
In my view this is the best strategy for winning.
I think people look at this wrong. It isn't about passing vs. rushing in terms of wins and losses. Both are needed. It isn't one or the other. The way it should be looked at is that points come out of the passing game and being able to control a game comes from rushing the ball. The Gianta run the ball because they know they aren't going to win in a shootout. They are running to control the game and keep it close. The Chiefs pass the shit out of the ball because they can and will take their chances in a shootout.
+1
I would never knock the analysis, I have too much respect for the expertise. I just don't feel that game state, how teams script their calls and adjust based on situations can/is captured.
I'm also not sure that every team that needs to run the ball as you suggest is necessarily good at building their run game. Somebody with a lot of experience in the league recently mentioned Bill Walsh's view that building the run game was the most difficult and important task of an offensive coach. Essentially, there were times when even the model west coast offense was going to need to run the ball. I think KC would be more effective if they wanted to be.
I just took a quick peek at the number of passing attempts for all 32 NFL teams as it relates to winning percentage. If I understand your statement correctly, we should find a negative correlation between the two sets of data. In other words, the more times a team passes the ball, the lower its winning percentage (and vice versa.) Bottom line, I found no correlation whatsoever between the two variables.
That doesn't mean you are wrong when considering a specific game. All of us have seen games when a team is ahead and kills the clock by running instead of passing. But in the context of all 32 NFL teams through 12 weeks of the season, there is no relationship between Passing Attempts and Winning Percentage.
Link - ( New Window )
Toward the end of the game, if Team A is winning, their goal may well be to just burn as much time off the clock as possible by running the ball. (No passing.) Note that Team A doesn't really care how many yards they gain on the ground... they just want to burn the clock.
Team B is losing to Team A, and they know that -- toward the end of the game -- Team A just wants to run the ball. So, Team B loads up its defense to stop the run.
In the scenario above, the winning side (Team A) will depress their average yards per carry since they are just trying to hold onto the ball, which would also include taking a knee. The losing side (Team B) will probably enhance their defensive stat against the run since they are loading up the box knowing full well that a running play has been called.
Bottom line: The winning team will have its Run Ratio diminished, whereas the losing team will have its Run Ratio enhanced.
the resultant equation is: Winning Percentage = (0.17312 * Run Ratio) + (0.69085 * Pass Ratio) - 0.42455
The coefficient for the Pass Ratio (.69085) has four-times the value of the Run Ratio coefficient (.17312).
When I plug each team's data into the regression equation, the three teams with a predicted winning percentage that is LOWER than its actual winning percentage are:
Minnesota
Giants
LA Chargers
the resultant equation is: Winning Percentage = (0.17312 * Run Ratio) + (0.69085 * Pass Ratio) - 0.42455
The coefficient for the Pass Ratio (.69085) has four-times the value of the Run Ratio coefficient (.17312).
When I plug each team's data into the regression equation, the three teams with a predicted winning percentage that is LOWER than its actual winning percentage are:
Minnesota
Giants
LA Chargers
I meant to say:
The three teams with the LOWEST predicted winning percentages relative to their actual winning percentages are:
Minnesota
Giants
LA Chargers
In other words, Minny, NY & Chargers are winning at a much higher rate than one would predict based on their respective Pass and Run Ratios.
I looked at the data a slightly different way. I removed the efficiency out of the running and passing statistics. In other words, rather than looking at offensive and defensive yardage on a per carry (or per attempted pass) basis, I just looked at rushing and passing yardage on a PER GAME basis without any reference to the number of times a team ran the ball or the number of times the team passed the ball.
Bottom line: The ratio "Rushing Yards per Game / Rushing Yards allowed per game" had about the same predictive power (of winning percentage) as the ratio "Passing Yards per Game / Passing Yards allowed per game." In fact, the rushing ratio per game had a slightly higher correlation with winning: 12.7% vs 9.4 % for the passing ratio.
So, there's something for everyone. If you are looking at "efficiency yardage" -- that is yards on a per attempt basis, then passing efficiency seems to have much more predictive power of winning than running efficiency. But if you are just looking at total yardage per game, then it's about even, with rushing having slightly more predictive power.
Taking a step back, what wins football games? Control the line of scrimmage, win the turnover battle would be my answer for the most impactful elements.
What statistics analysis might support that outlook? Maybe yards per play versus yards per play allowed? Maybe total plays versus total plays allowed? Turnovers committed versus allowed?
I would expect to see a huge correlation between winning and total plays versus total plays allowed? What does it all mean….honestly, probably not much. Get the horses in the trenches and the QB and the stats will be in your favor however you look at them.
I don't think it's as cut and dried as you started with.
Or not.
A regression/ correlational analysis on football stats cannot tell a coach (or his players) what they should do to play winning football.
Nor can it say which stat leads to winning football since - to quote a very old statistical saying - “correlation is not causality.”
What regression can do is plow through a mass a data to reveal underlying patterns in the data. If the pattern seems to fit everyone’s intuition, well then everything is fine. But when a data pattern does not fit pre-conceived notions, it forces us to either figure out “why” and/or modify our pre-conceived notions.
Or not.
A regression/ correlational analysis on football stats cannot tell a coach (or his players) what they should do to play winning football.
Nor can it say which stat leads to winning football since - to quote a very old statistical saying - “correlation is not causality.”
What regression can do is plow through a mass a data to reveal underlying patterns in the data. If the pattern seems to fit everyone’s intuition, well then everything is fine. But when a data pattern does not fit pre-conceived notions, it forces us to either figure out “why” and/or modify our pre-conceived notions.
What did you glean from the data patterns you found?
Quote:
Or not.
A regression/ correlational analysis on football stats cannot tell a coach (or his players) what they should do to play winning football.
Nor can it say which stat leads to winning football since - to quote a very old statistical saying - “correlation is not causality.”
What regression can do is plow through a mass a data to reveal underlying patterns in the data. If the pattern seems to fit everyone’s intuition, well then everything is fine. But when a data pattern does not fit pre-conceived notions, it forces us to either figure out “why” and/or modify our pre-conceived notions.
What did you glean from the data patterns you found?
My #1 take-away is that, while we often judge a RB by his yards per carry, the statistic on a total team-wise basis shows no pattern (and is not predictive) with winning percentage. Below is an explanation why that may be so**.
**Here's one thought on why the Run Ratio
M.S. : 11:01 am : link : reply
appears to have no correlation with winning percentage.
Toward the end of the game, if Team A is winning, their goal may well be to just burn as much time off the clock as possible by running the ball. (No passing.) Note that Team A doesn't really care how many yards they gain on the ground... they just want to burn the clock.
Team B is losing to Team A, and they know that -- toward the end of the game -- Team A just wants to run the ball. So, Team B loads up its defense to stop the run.
In the scenario above, the winning side (Team A) will depress their average yards per carry since they are just trying to hold onto the ball, which would also include taking a knee. The losing side (Team B) will probably enhance their defensive stat against the run since they are loading up the box knowing full well that a running play has been called.
Bottom line: The winning team will have its Run Ratio diminished, whereas the losing team will have its Run Ratio enhanced.