## Crunching The Numbers: Under Further Review (Part III of IV)

Mark J. Rebilas-USA TODAY Sports

The previous post discovered that some statistics don't correlate with winning - at all. This prompted the next leg of my journey, where I analyzed other statistics (a LOT of them) to find out which ones really do correlate to winning. The results may surprise you.

At the end of the Part II, the regression analysis showed some interesting correlations between statistics and winning. Namely, the yards per carry metric is apparently meaningless. With a result this surprising, the only reasonable thing to do was to perform a regression analysis on more statistics in an attempt to draw some concrete conclusions. The statistics I use for my algorithm come from Team Rankings (it's an awesome site if you're a stat junkie), so I went there. As a disclaimer, I stayed with statistics in the facets I was measuring in my algorithm (running game, passing game, turnovers, and scoring), which means I left out statistics related to special teams and penalties. While that indicates this analysis is by no means exhaustive, I did analyze seventy-three individual metrics for all thirty-two teams, so it is fairly comprehensive.

In the interests of getting to the point, the tables below will only show the statistics that had a 95% confidence level - if you recall from Part II, this means their correlation coefficient has an absolute value greater than 0.3494. At the end of the post I did include a table of every metric that I analyzed should you feel inclined to look them over and draw your own conclusions.

And if you're just joining us now, check out how I got started with Part I and an explanation of regression analysis can be found in Part II. Now, onto the numbers!

#### OFFENSE

To start, let's look at the offense as a unit.

 OFFENSE OVERALL Statistic Correlation Points Scored/Game 0.7050 Fourth Quarter Scoring 0.6207 Third Quarter Scoring 0.5748 Second Quarter Scoring 0.4977 Third Down Efficiency 0.3555

Nothing too earth-shattering here. The one noticeable trend is that teams that won tended to score more in the second half than in the first half (in fact, first quarter didn't even make the cut with its confidence level). While this is interesting for sure, the "score-by-quarter" metric is one that I think has the potential to be volatile. It's something I'll be paying attention to across seasons to see whether or not this trend holds or if this was merely a coincidence. Moving on to the running game:

 RUSHING OFFENSE Statistic Correlation Rushing First Downs/Game 0.4542 Rushing Play % 0.4243 Rushing Attempts/Game 0.4184

This has much more to ponder. Remember how the yards-per-carry statistic (which essentially measures the effectiveness of your running game) was meaningless? All of the meaningful run statistics are bulk statistics. That is, they are all measurements of how often a team runs the ball, not how well a team runs the ball. I found the first down statistic to be particularly interesting. At first I chalked it up to a false correlation - it had more to do with teams that got more first downs overall being more successful than it had to do with the running game. But when I looked at the correlation for passing first downs it was only 0.22. So the fact that they are rushing first downs seems to have more of an effect. Speaking of the passing game, which statistics mattered there?

 PASSING OFFENSE Statistic Correlation Average Passer Rating 0.6294 Yards/Pass Attempt 0.6099 Incompletions/Game -0.5405 Yards/Completion 0.5392 Passing Play % -0.4243 Completion Percentage 0.4197 QB Sacked/Game -0.3989

There's a lot going on here. If you look over the statistics that showed up here, the passing game takes on the opposite characterization from the running game. The important metrics here for the most part measure how well the ball is thrown, not how often it's thrown. In fact, the passing play percentage metric has a negative correlation with winning, meaning that losing teams threw the ball more often. Considering this, Andy Reid's success in Philadelphia sounds miraculous given his allergy to handing the ball off. Jokes aside, part of this negative correlation might come from the idea that bad teams pass more because they're trying to dig themselves out of holes. It's hard to say for sure, especially when reviewing the other statistics up there. For example, there is a strong negative correlation with incompletions... but no corresponding strong positive connection with completions. This apparent trend that passing the ball is more about efficiency than anything else is something I'll touch up on later.

#### DEFENSE

Of course, football isn't just about offense. There's a reason why the phrase, "Defense wins championships" exists. Take a gander at some of the numbers for the defensive side of the ball:

 DEFENSE OVERALL Statistic Correlation Points Allowed/Game -0.7617 Second Quarter Scoring -0.6299 Fourth Quarter Scoring -0.5320 First Quarter Scoring -0.4635 Third Quarter Scoring -0.3983

Just to reiterate, the negative sign indicates that as the values for the metrics decrease, wins increase. This makes sense for a lot of defensive stats, which is why there are more negative coefficients on this side of the ball. There is a much stronger correlation here with points than the offense had, with all four quarters being represented (although third down defense is noticeably absent). Interestingly enough the trend with score-by-quarters does not really exist here, which reinforces my suspicion that it might be a coincidence. The only possible connection I can think of would be the idea of "stomping out momentum" that the opposing offense might be generating at the end of each half. Again, the more likely explanation here is simply coincidence. Now, how do run defenses fare?

 RUSHING DEFENSE Statistic (Opponents) Correlation Rushing Play % -0.5186 Rushing Attempts/Game -0.4693

The message here is pretty clear, and it's a tall task for any football team: don't stop the run; force your opponent to abandon it entirely. Logically speaking, if you're unable to stop the run your opponent is unlikely to quit handing the ball of, but we've previously seen how little yards per carry allowed has to do with winning. This means some of the onus falls on the offense to build a lead and force the opponent to play catch-up with the passing game... but at the same time this contradicts the established trend that teams who score later in the game win more often. So what is the bottom line here? I'm going to use this as reinforcement that the score-per-quarter trend is a coincidence since more evidence exists to support the "more running means more winning" concept. Does a similar "less passing means more winning" trend exist with the passing game?

 PASSING DEFENSE Statistic (Opponents) Correlation Average Passer Rating -0.6205 Passing Play % 0.5215 QB Sacked/Game 0.4758 Incompletions/Game 0.4244 Yards/Pass Attempt -0.4233

THE TREND SAYS THAT THE MORE TEAMS PASS, THE MORE THEY LOSE.

The short answer is, "emphatically yes." Look at that correlation for passing play percentage. That means the more your opponent passed the ball, the more games you won. The other statistics refer back to the "passing efficiency trend" I mentioned in Part II; they're all associated with disrupting the quarterback's ability to perform. Sacks, yards per attempt, incompletions, passer rating... they all point to making sure the quarterback is ineffective. And when combined with the passing play percentage, it makes sense. If you can force poor quarterback play and then force the opponent to abandon the other dimension to their offense in relief, you get a pretty simple formula for success.

#### TURNOVERS

Turnovers: the magic buzz word of the NFL. It's impossible to turn on any football analysis show without some pundit mentioning the importance of turnovers. But there is a lot of data buried underneath that word. How often did your offense fumble? How often did they recover those? What's more valuable, interceptions or interception percentage? Numbers don't always have the answer, but in this case they do:

 TURNOVERS Statistic Correlation Turnover Margin/Game 0.6560 Interceptions/Game -0.5353 Giveaways/Game -0.5289 Interception % -0.4821 Takeaways/Game 0.4541 Statistic (Opponents) Correlation Interceptions/Game 0.4797 Interception % 0.4554

The first noticeable observation is that turnovers seem to focus on offensive performance more than defensive performance. I'm not very surprised that fumbles don't really show up here; they tend to be more random than interceptions and a lot of them are based on luck. Interceptions have their fare share of luck too, but a good coach can disguise coverages and apply pressure to induce a pick. The same really can't be said for fumbles. It's also interesting how turnover success is more dependent on a team's ability to take care of the football over its ability to take it away from their opponent. I would say this makes sense; by mitigating error a team is increasing the odds that the opponent will make more than them. It's also worth noting that other than scoring-related statistics, turnover margin had the highest correlation to winning than any other metric I analyzed.

#### The Verdict (Part III)

THE NFL IS NOT A PASSING LEAGUE - IT'S A QUARTERBACK'S LEAGUE.

After all of these numbers, what conclusions can we draw? To put it bluntly, the NFL is not a passing league. It is a quarterback's league. There is a distinct difference here. It's not about filling the air with footballs; if anything, the evidence suggests this is how you lose. It's about giving your quarterback an environment to succeed and then having one that can execute consistently and efficiently. Success happens when a team complements this with a healthy dose of the running game (preferably more than fifty percent). It doesn't even matter how well the running game is working - the key is to just keep handing it off. Likewise, defenses thrive by having the ability to compromise the play of the opposing quarterback and forcing an abandonment of the running game, which is a potent combination.

So now, the big question, the one I know we all care about: how did the Eagles fare in these metrics? I look at exactly that in the conclusion to this series in Part IV.

#### For Reference

Should you have some extra time on your hands and an interest in this topic, the chart below documents every statistic that I analyzed:

 OFFENSE OVERALL Statistic Correlation Third Down Efficiency 0.3555 Points Scored/Game 0.7050 First Quarter Scoring 0.3248 Second Quarter Scoring 0.4977 Third Quarter Scoring 0.5748 Fourth Quarter Scoring 0.6207 DEFENSE OVERALL Statistic Correlation Third Down Efficiency -0.3402 Points Allowed/Game -0.7617 First Quarter Scoring -0.4635 Second Quarter Scoring -0.6299 Third Quarter Scoring -0.3983 Fourth Quarter Scoring -0.5320 RUSHING OFFENSE Statistic Correlation Rushing Attempts/Game 0.4184 Rushing Yards/Game 0.3028 Rushing First Downs/Game 0.4542 Yards/Rush Attempt 0.0517 Rushing Play % 0.4243 Rushing First Down % 0.2296 Rushing Yards % 0.1206 PASSING OFFENSE Statistic Correlation Pass Attempts/Game -0.2741 Completions/Game -0.0333 Incompletions/Game -0.5405 Completion Percentage 0.4197 Passing Yards/Game 0.2207 Yards/Pass Attempt 0.6099 Yards/Completion 0.5392 QB Sacked/Game -0.3989 QB Sacked % -0.1267 Passing First Downs/Game 0.2161 Passing First Down % -0.2194 Average Team Passer Rating 0.6294 Passing Play % -0.4243 Passing Yards % -0.1206 RUSHING DEFENSE Statistic (Opponents) Correlation Rushing Attempts/Game -0.4693 Rushing Yards/Game -0.2575 Rushing First Downs/Game -0.2559 Yards/Rush Attempt -0.0087 Rushing Play % -0.5186 Rushing First Down % -0.1728 Rushing Yards % -0.1236 PASSING DEFENSE Statistic (Opponents) Correlation Pass Attempts/Game 0.2556 Completions/Game -0.0414 Incompletions/Game 0.4244 Completion Percentage -0.3954 Passing Yards/Game -0.1365 Yards/Pass Attempt -0.4233 Yards/Completion -0.1752 QB Sacked/Game 0.4758 QB Sacked % 0.3410 Passing First Downs/Game -0.1087 Passing First Downs % 0.3055 Average Team Passer Rating -0.6205 Passing Play % 0.5215 Passing Yards % 0.1236 TURNOVERS Statistic Correlation Interceptions/Game -0.5353 Fumbles/Game -0.1820 Fumbles Lost/Game -0.2127 Fumbles Not Lost/Game -0.1685 Giveaways/Game -0.5289 Takeaways/Game 0.4541 Turnover Margin/Game 0.6560 Interception % -0.4821 Fumble Recovery % 0.0933 Giveaway Recovery % 0.0371 Takeaway Recovery % 0.1138 Statistic (Opponents) Correlation Interceptions/Game 0.4797 Fumbles/Game 0.1609 Fumbles Lost/Game 0.2995 Fumbles Not Lost/Game 0.0127 Interception % 0.4554 Fumble Recovery % -0.0933 Giveaway Recovery % -0.1139 Takeaway Recovery % -0.0371

## Trending Discussions

Log In Sign Up

forgot?
Log In Sign Up

### Please choose a new SB Nation username and password

As part of the new SB Nation launch, prior users will need to choose a permanent username, along with a new password.

Your username will be used to login to SB Nation going forward.

I already have a Vox Media account!

### Verify Vox Media account

Please login to your Vox Media account. This account will be linked to your previously existing Eater account.

### Please choose a new SB Nation username and password

As part of the new SB Nation launch, prior MT authors will need to choose a new username and password.

Your username will be used to login to SB Nation going forward.

### Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

### Forgot password?

Try another email?

### Almost done,

By becoming a registered user, you are also agreeing to our Terms and confirming that you have read our Privacy Policy.

### Join Bleeding Green Nation

You must be a member of Bleeding Green Nation to participate.

We have our own Community Guidelines at Bleeding Green Nation. You should read them.

### Join Bleeding Green Nation

You must be a member of Bleeding Green Nation to participate.

We have our own Community Guidelines at Bleeding Green Nation. You should read them.

### Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.