All offseason coaches talk about what they want their offensive identity to be. Most of the time it sounds something like what Chuck Pagano has been preaching since he arrived in Indianapolis. “You have to be able to run the football and stop the run”. Every week coaches also talk about scripting their first handful of plays. However, as Mike Tyson would say, “Everybody has a plan until you get punched in the face”.
What happens after the game starts is the game script and it greatly affects what plays the coaches call. To find out what the coaches true offensive identity is we need to peel away all the punches to the face that built the game script and see what they really wanted their identity to be.
To do this I obtained play by play data for every game since 2000 and tested several key variables that affected the coach’s decision to pass or run on every single play. Of these variables there were 10 that were very key and included:
- Score differential
- Minutes Left
- Last two minutes of the half
- Minutes Used * Score Difference (MUSD)
- Yards to go for first down
- Down and Distance
- Yards from Own Goal
- Vegas point spread prior to the game
- Wind speed
- Opponents defensive pass identity
In this article I will take a closer look at each of these key variables and how they helped shape each team’s pass ratio in 2013 and how it hid their true offensive identity. In part two, I will show how we can use these true offensive identities and some of these same variables to help project the pass ratios of each team this year.
The variable that is most influential on each team's pass ratio is simply the score of the game. As most of us know by now as team's get further behind they tend to pass more often in order to quickly catch up and when they start to build a lead they tend to run more often to grind out the clock. But did you realize how strong the correlation really was? The graph below shows how often teams pass when leading or trailing by a specific score along with the best fit line for each possible score all the way up to 28 points which is when the game really starts to get out of hand. Amazingly the R squared is .93 which means that 93% of the variance in pass ratio is explained simply by looking at the score of the game. Over a large enough sample all you really need is the score of the game to help you predict how often each team will pass. Talk about letting the punch to the face change your plan.
The next variable that we will look at is time left in the game. First let’s break this down into quarters then look at it on a minute by minute basis.
| Quarter | Pass Ratio |
| 1 | 52.5% |
| 2 | 59.7% |
| 3 | 55.0% |
| 4 | 57.2% |
| OT | 51.5% |
As you can see, coaches really do try to start off by running the ball as they pass the ball only 52.5% in the first quarter but by the end of the game they are now passing at a 57.2% rate. However, what might stand out as unusual is the fact that the second quarter actually has the highest pass ratio. But if we look at the chart below not only do we see that coaches are passing more as the game increases we see the huge spike that occurs during the last two minutes of the second half where approximately 72% of all plays are pass attempts as teams try to quickly march down the field to get that last second scoring opportunity. A big part of your quarterback and wide receiver’s weekly scores are going to depend on if they got the ball in the last two minutes of the half. For example, Denver led the league last year with 72 plays in the last two minutes and they passed on 55 of them for a whopping 77% pass ratio. It’s no surprise that Peyton Manning lead the league in fantasy points.
Of course, what is really important is the interaction between time left and score differential. Being behind 7-0 with 12 minutes left in the first quarter is one thing but being down 24-17 with a minute left in the fourth quarter is something completely different. What I did to measure this interaction is multiply the minutes used in the game by the score differential to get the variable I am calling “Minutes Used * Score Differential” (MUSD). For example, the first example above would be equal to a score of -21 (3 minutes use * -7 score difference) for the trailing team where the second example would be equal to -413 (59 * -7). This number is essentially estimating your chances of winning, the higher the number the better your chances are of winning. I tried to use a win probability metric instead of this calculation but it actually faired a bit worse. This may be because coaches only have so much time to think between play calls and they think in terms of time and score not odds of winning. The R-Squared of MUSD is .94 and ends up being the most important variable in determining the play calling.
We have now seen that score differential and time remaining have a huge impact on the team’s play calling but so does the down and distance of the play and where the ball lies on the field. Let’s start by looking at the pass ratios by down alone. These ratios follow the same basic premise as the score data we saw above. Coaches want to run when they have a chance like on first and second down. On both of these down they actually pass and run about equally. However, when the situation becomes a bit more ominous like third down they start passing very aggressively. In fact, on third down they start passing at a rate of 76.3% which is one of the highest rates we have seen yet.
| Down | Pass Ratio |
| 1 | 47.5% |
| 2 | 54.8% |
| 3 | 76.3% |
| 4 | 60.5% |
What happens as we start to look at the ratios by yards to go? Here we get a much more logarithmic looking chart where we see a steady decline in the pass ratio as the yards to go decreases until we get to about 2 yards to go where it drops like a rock and coaches start to favor running. The other thing that sticks out is the significant drops at 10, 15, and 20 yards to go. But, if we start thinking about what plays happen most often at these yards to go you quickly realize that most of these plays are occurring on first down. If we exclude first downs, as I did in the following chart, you can see this clears up all of the unusualness at the 15 yard line. However, the play calling at the 10 yard line is still out of line with the other nearby yard lines. This is probably because in order to get to 2nd-and-10 that means the team probably passed on first down and they don’t want a chance at three straight incomplete passes as they will be forced to pass on third down too.
To better see this interaction between down and distance let’s look at the pass ratio by each down and distance. Here the sample sizes are getting a lot smaller so they are a bit noisier, especially on first and fourth down, but we can still get an idea of how teams are calling plays on each down.
Here the first thing that stands out to me is that if it is third or fourth down coaches are really predictable. If it’s one yard to go they are going to run it, if it’s more than 3 yards to go they are going to pass. Only when it is 3rd-and-2 are coaches truly unpredictable. The other thing that sticks out to me is how low the pass ratios are for 1st and anything less than 10 yards. This is driven primarily by the fact that most of the plays of this distance are ran near the opponents end zone which leads us to my next graph which shows the pass ratio by yards from the team’s own goal.
The main focus of this graph is how often teams run if they are within a few yards of scoring or if they are backed up at their own goal line. The rest of the time the pass ratio is pretty consistent in the 55%-60% range. Here again we also see the coaches trying to run the ball as often as possible by the dips at the 20 and 30 yard line which is where teams take over drives after a touchback (or a touchback and a penalty). Even these dips still fall within this range though.
The last two factors I want to focus in on are both known before game time and will be key factors in predicting next year’s pass ratios as well as pass ratios each week. They are the Vegas point spread and wind speed at game time.
As we can see in both cases there is a strong linear correlation between the pass ratio and the variable. In fact, the wind speed has an R-Squared of .48 and the Vegas Line has an R-Squared of .52 which are pretty strong for variables that are known well before the plays are even close to being ran.
Using the various odds obtained above I determined what the odds were of passing for every play last year and then compared those odds with what really happened. This allowed me to see what teams really wanted their pass identity to be and subsequently how often the game situation got in the way. I then did this for the defenses as well and iterated the results between the defense and offense.
The table below lists all teams and is sorted by their offensive identity. Offensive identity is calculated by simply taking the actual pass attempts divided by the expected pass attempts. The table below that lists each team’s defensive identity and is laid out in the same format. Here’s how to read the tables. The Dallas Cowboys passed on 65% of their plays last year. Based on the game situations mentioned above we would have expected the Cowboys to pass only 57.1% of the time. This led to an offensive identity that was 114% of the league average. Since the league average pass ratio last year was 58.3% if the Cowboys faced league neutral situational plays they would have passed at a situational neutral pass ratio of 66.2% which was 2% lower than what they passed at. (Note these stats do include the playoffs and week 17 and quarterback scrambles are counted as runs)
| Team | Actual Pass Ratio | Expected Pass Ratio | Offensive Identity | League Avg Ratio | Situation Neutral Pass ratio | Difference |
| DAL | 64.9% | 57.2% | 114% | 58.3% | 66.1% | -2% |
| CLE | 67.9% | 59.9% | 113% | 58.3% | 66.0% | 3% |
| DEN | 60.6% | 53.5% | 113% | 58.3% | 66.0% | -8% |
| ATL | 68.7% | 60.6% | 113% | 58.3% | 66.0% | 4% |
| MIA | 65.3% | 58.4% | 112% | 58.3% | 65.2% | 0% |
| NO | 62.9% | 56.4% | 111% | 58.3% | 65.0% | -3% |
| PIT | 61.5% | 57.5% | 107% | 58.3% | 62.3% | -1% |
| KC | 57.3% | 53.9% | 106% | 58.3% | 62.0% | -7% |
| IND | 61.2% | 58.1% | 105% | 58.3% | 61.4% | 0% |
| BAL | 61.2% | 58.9% | 104% | 58.3% | 60.6% | 1% |
| DET | 59.7% | 58.1% | 103% | 58.3% | 59.8% | 0% |
| ARI | 59.3% | 58.4% | 102% | 58.3% | 59.2% | 0% |
| CIN | 57.0% | 56.3% | 101% | 58.3% | 59.0% | -3% |
| NE | 57.9% | 56.6% | 102% | 58.3% | 59.6% | -3% |
| NYG | 61.5% | 61.1% | 101% | 58.3% | 58.6% | 5% |
| CHI | 60.3% | 60.1% | 100% | 58.3% | 58.4% | 3% |
| GB | 56.8% | 57.6% | 99% | 58.3% | 57.5% | -1% |
| HOU | 62.0% | 62.6% | 99% | 58.3% | 57.7% | 7% |
| Jax | 62.9% | 64.6% | 97% | 58.3% | 56.7% | 11% |
| SD | 53.4% | 56.1% | 95% | 58.3% | 55.5% | -4% |
| CAR | 51.9% | 54.7% | 95% | 58.3% | 55.3% | -6% |
| PHI | 53.0% | 56.6% | 94% | 58.3% | 54.6% | -3% |
| MIN | 58.2% | 62.6% | 93% | 58.3% | 54.2% | 7% |
| TB | 57.2% | 60.5% | 95% | 58.3% | 55.1% | 4% |
| STL | 56.0% | 59.6% | 94% | 58.3% | 54.8% | 2% |
| WAS | 59.0% | 63.9% | 92% | 58.3% | 53.8% | 10% |
| TEN | 55.2% | 60.2% | 92% | 58.3% | 53.5% | 3% |
| OAK | 56.5% | 62.8% | 90% | 58.3% | 52.4% | 8% |
| SF | 47.8% | 53.3% | 90% | 58.3% | 52.2% | -9% |
| SEA | 47.3% | 52.7% | 90% | 58.3% | 52.3% | -10% |
| BUF | 51.0% | 57.6% | 89% | 58.3% | 51.6% | -1% |
| NYJ | 51.6% | 61.6% | 84% | 58.3% | 48.8% | 6% |
Here is the same table but for the defensive side of the ball.
| Team | Actual Pass Ratio | Expected Pass Ratio | Defensive Identity | League Avg Ratio | Situation Neutral Pass ratio | Difference |
| MIN | 61.0% | 56.3% | 109% | 58.3% | 63.2% | -3% |
| NYJ | 59.8% | 55.2% | 108% | 58.3% | 63.2% | -5% |
| ARI | 64.5% | 59.6% | 108% | 58.3% | 63.0% | 2% |
| WAS | 55.4% | 52.2% | 106% | 58.3% | 61.8% | -10% |
| SD | 60.7% | 57.8% | 105% | 58.3% | 61.1% | -1% |
| NYG | 58.4% | 56.3% | 104% | 58.3% | 60.4% | -3% |
| CIN | 61.3% | 59.8% | 103% | 58.3% | 59.7% | 3% |
| DET | 61.6% | 60.5% | 102% | 58.3% | 59.3% | 4% |
| CAR | 62.8% | 61.8% | 102% | 58.3% | 59.2% | 6% |
| CLE | 58.3% | 57.4% | 102% | 58.3% | 59.1% | -1% |
| PIT | 58.3% | 57.5% | 101% | 58.3% | 59.1% | -1% |
| PHI | 60.8% | 60.1% | 101% | 58.3% | 59.0% | 3% |
| OAK | 57.1% | 56.9% | 100% | 58.3% | 58.5% | -2% |
| DAL | 60.1% | 60.0% | 100% | 58.3% | 58.3% | 3% |
| TEN | 56.6% | 56.5% | 100% | 58.3% | 58.3% | -3% |
| BAL | 57.5% | 57.5% | 100% | 58.3% | 58.2% | -1% |
| GB | 57.1% | 57.3% | 100% | 58.3% | 58.1% | -2% |
| jax | 53.5% | 54.1% | 99% | 58.3% | 57.7% | -7% |
| DEN | 60.9% | 61.7% | 99% | 58.3% | 57.5% | 6% |
| TB | 57.0% | 58.0% | 98% | 58.3% | 57.3% | -1% |
| MIA | 56.2% | 57.5% | 98% | 58.3% | 57.0% | -1% |
| BUF | 56.6% | 58.0% | 98% | 58.3% | 56.9% | -1% |
| NE | 57.8% | 59.4% | 97% | 58.3% | 56.8% | 2% |
| KC | 60.6% | 62.8% | 96% | 58.3% | 56.2% | 8% |
| STL | 56.7% | 58.9% | 96% | 58.3% | 56.1% | 1% |
| ATL | 54.6% | 57.0% | 96% | 58.3% | 55.9% | -2% |
| IND | 55.4% | 57.8% | 96% | 58.3% | 55.8% | -1% |
| HOU | 53.2% | 55.6% | 96% | 58.3% | 55.8% | -5% |
| CHI | 52.8% | 55.2% | 96% | 58.3% | 55.7% | -5% |
| SF | 59.8% | 62.6% | 96% | 58.3% | 55.7% | 7% |
| NO | 58.0% | 60.7% | 96% | 58.3% | 55.7% | 4% |
| SEA | 58.6% | 63.8% | 92% | 58.3% | 53.5% |
9% |
What do these ratios mean to you? Does the order represent how you would think of each team in terms of offensive identity? Hit me up on twitter @stevebuzzard to discuss. In part two, I will share my own thoughts on these identities and propose how you can use these numbers to project this year’s pass ratios.
