Football is back! And with football comes Crunching The Numbers, my annual investigation to find the statistics that serve as “indicators” of success. For those of you who may be new to the site, this project has been ongoing since 2011, when I became interested in why some teams start out hot and then fade, or vice versa. Where there underlying factors that could have predicted the drastic change in performance? Beyond the “duh” statistics - points, turnovers, etc - what other metrics can we connect to winning?
My analysis has taken many forms over the years, but the presentation I like the most was the one I started last year, where I use my metrics to preview the Eagles’ upcoming opponents, offering things to watch out for that may not be obvious at first. Based on some of your feedback, I’m going to make a concerted effort to make these previews a bit less dry than they were last year. The first preview will be for the Vikings game, as I like to wait four weeks in order to have enough data to analyze.
Before diving into the actual statistics I’ll be using, let’s briefly discuss how I chose them.
Generally speaking, I choose statistics by first theorizing what I think it takes for a team to be successful in the NFL, then by finding statistics that correspond to these theories, and finally testing these statistics by correlating them to wins. (Note - a few of you have suggested that I correlate to point differential over wins. I am intrigued by this, and it will be something I will investigate for 2019. I had unfortunately completed my analysis before that suggestion was made.) The results of the test will either validate or disprove my theories, and I will draw conclusions from there.
For my theories, I like to keep them in broad strokes, because this makes them easier to test. A statement like, “Teams with a positive turnover ratio are more likely to win” is easier to prove or disprove than, “Teams that turn the ball over late in games lose more than teams that turn the ball over early in games.” It’s difficult to draw conclusions from statements like that in a vacuum, because of course football does not take place in a vacuum.
All of that being said, here are my general theories to winning in professional football:
- The foundation to winning on offense is a balanced approach with a heavy emphasis on being successful in the passing game. My personal analogy for this is that if your offense is the engine of a car, the passing game is the motor while the running game is the oil. The oil only serves a complementary role to the motor, which is what moves the car. But if you don’t have any oil at all, the engine will eventually seize and the car will stop. Likewise, the passing game is what will score you points - but if you never hand the ball off, defenses will adapt and find a way to stop you. It is not necessary for those handoffs to be successful, however. They just need to happen.
- Scoring earlier in the game is more important than scoring later. If you can jump out to a lead, you’ll force the other team to be one-dimensional on offense (see #1 above), allowing the defense to pin their ears back and get after the quarterback.
- On defense, the emphasis should be on the passing game, like it is with the offense. Having a porous run defense is not desirable, but it’s the lesser of two evils.
- Time of possession matters. We all remember when Chip Kelly scoffed at the notion of time of possession, but in the NFL with 53-man rosters, keeping your defense fresh while limiting the number of opportunities the opponent has to score is more important than any advantage gained by wearing the other team down with tempo.
I’ll use these theories to pick statistics I want to test. From there, I’ll collect the past three years of season averages for each team and use a simple linear regression to correlate with their win percentage over that time frame. If you’re asking why I only go back three years, I’m trying to stay current with the ever-changing NFL and I feel that three years is enough to pick out trends without getting noise from other play styles that may no longer be in vogue.
Here are statistics I analyzed (taken from Team Rankings), ordered from highest correlation to lowest correlation. The ones in bold face are my selections for this season’s Crunching The Numbers. Remember, for a sample size of 32, the magic numbers are 0.35 and -0.35, which gives us a 95% level confidence in the significance of the relationship and helps us to avoid the oft-dreaded “correlation does not imply causation” fallacy:
- Yards Per Point (Y/PT): -0.865
- Points/First Half (PTS/1HLF): 0.720
- Points/Second Half (PTS/2HLF): 0.717
- Time Of Possession (TOP): 0.627
- Yards Per Pass Attempt (Y/PA): 0.519
- Sack Percentage (SACK%): 0.365
- Offensive Sack Percentage (OFFSCK%): -0.301
- Pressure Rate (PRESS%): 0.204
- Rushing Play Percentage (RUSH%): 0.204
- Yards Per Rush Attempt Allowed (YPRA): -0.117
- Yards Per Rush Attempt (Y/RA): -0.090
- My theory about scoring early versus scoring late does not seem to hold any water. Over three years, a difference in correlation of 0.003 cannot be chalked up to anything significant. Scoring is just scoring, regardless of when it occurs.
- I wrote in depth about sacks here, so it was intriguing that they have produced a stronger correlation to winning over the past three years. For now, they will be included in Crunching The Numbers, but I’ll have to keep re-checking going forward. Strangely enough, the correlation with preventing sacks is still somewhat strong, but not as much.
- It really cannot be understated how little connection there is between running efficiency and winning. This is a topic that has been covered on several football analytics sites, including Football Outsiders and Cold Hard Football Facts. The correlation between yards per rush and winning was actually slightly negative, meaning that over the past three seasons teams with worse running efficiency actually won more often, although the connection is so weak that it’s probably just a coincidence. Stopping the run and keeping a balanced offense (rushing percentage) both had stronger correlations, but still not enough to be significant, and both pale in comparison to the very strong connection we see between winning and yards per pass attempt.
- The insignificance of running efficiency is probably linked to how difficult it is to generate explosive plays (generally defined as plays of 20+ yards) by running versus passing. This bears out with our strongest metric, yards per point. It has a strong negative correlation, which makes sense: if you need fewer yards to score points, your offense is running more efficiently than your opponent. This is even more interesting when you observe the strong correlation to time of possession, which suggests that it’s not necessarily about having more possessions than your opponent, but is instead about deliberately controlling the clock by pacing the offense and milking the play clock.
Given this information the statistics I’ll be using this season for Crunching The Numbers are yards per point, time of possession, yards per pass attempt, and sack percentage. Keep your eyes open for Eagles game previews using these metrics come Week 5, where I’ll start offering some things to look for as they gear up to take on the Vikings!
What do you think? Can these metrics tell us things we don’t already know? Chime in with your questions, suggestions, and snarky retorts in the comments below!