It’s a passing league – are teams really better off with a balanced offense?
This is an analysis I have been thinking about for a while and even more recently with the debate on the Eagles run / pass balance. Usually I have an idea where an analysis will land, but going into this I really did not know what to expect. I probably expected passing-heavy drives to be more successful but the data shows that successful drives are more likely to be balanced than passing heavy.
I am not a “smash mouth, old school, let’s run all day” believer, but I also differ with a common view in the analytics community that passing is always better than running. Expected Points Added (EPA), the key metric used to analyze players, plays, and games, is almost always negative for runs and positive for passes. That means that on any given play, regardless of down, distance, point in the game, etc., a pass has more expected value than a run.
For 2021, EPA splits for running and passing are the following:
- Passing: League average +0.173 EPA per play with 29 of 32 teams averaging a positive EPA on passing attempts
- Rushing: League average -0.039 EPA per play with only 8 teams having a positive EPA for rushing attempts
(The 8 teams with positive rushing EPA this season are Arizona, Baltimore, Buffalo, Cleveland, Indy, Kansas City, Philly, and Tennessee… interesting to note that 6 of the 8 are teams with top rushing QBs)
Run/pass balance for the league and Eagles
A couple of points on the data used. The dataset I am using is from 2015-2021 and calculates at a drive level using the number of passes and runs for each drive. It ended up not making much of a difference in the overall data but I excluded three types of drives:
- “Garbage time” drives where win expectancy was over 90%
- Drives with turnovers
- Drives with penalties
From 2015-2021, the league (blue bar) had a median drive balance of 66% passing / 34% rushing.
The Eagles (red bar) had the same median drive balance although most in Philly, including me, would not believe that. Philly does skew towards heavier passing though – the top red box shows the centerline (median) at 66% but the right-hand side of the box stretches out to 80% passing, meaning that 25% of their drives were between 66-80% passing (vs. the league average where 25% of drives were between 66-75%). Not a huge skew, but the Eagles definitely use more passing-heavy drives.
Value of drives by run/pass balance
Below I have grouped drives into quintiles by run/pass balance – the far left group includes drives that used between 0-20% passing, the second group is drives between 20-40% passing, and so on. To the left are run-heavy drives and to the right are pass-heavy drives.
Excess Points scored by run/pass balance
The first chart shows excess points scored by drive run/pass balance grouping. Excess points are what was actually scored on a drive vs. the expected points at the beginning of the drive – all drives have an expected point value at the start of the drive based on starting field position, time left in the half, and score differential. I like excess points as a metric here as it normalizes shorter and longer drives.
The middle quintile between 40-60% passing (“balanced drives”) has the highest average points scored vs. expected drive points with a very slight positive 0.02 excess points scored. It also has the highest range with 25% of balanced drives scoring between +0.02 and +3.93 excess points. All other quintiles have negative median excess points.
Drive EPA by run/pass balance
This next look uses average drive EPA instead of excess points, but has the same result with the middle quintile (“balanced drives” between 40-60% passing) having the highest average drive EPA and only quintile with a positive median drive EPA.
And if you look at drives not grouped into quintiles, it is messier but shows the same overall point with the trendline of excess points highest on drives between 50-55% passing.
The two ends – 0% passing drives and 100% passing drives – both have mini-peaks as well which was puzzling. It turns out these drives are primarily drives that start deep in the opponent territory (the opponent’s 30 yard line or lower).
Drives ending in score by run/pass balance
If you look at the result of the drive – whether the drive resulted in a score or not – again the same point is made. Drives ending in a touchdown averaged 62.5% passing, drives ending in a field goal averaged 63.4% passing, and drives that did not score averaged 66.7% passing. Again, these filter out garbage time. Not huge differences in run/pass balance, but again drives that lean more balanced have higher success rates.
2021 team run/pass balance and value
And lastly, a look at every team’s run/pass balance vs. excess points scored in 2021. Teams with better QBs have higher excess points scored (no surprise) but they are not all heavy passing teams – Brady, Rodgers, Dak, and Herbert are all more pass-heavy than the league average but Kyler, Tannehill, Josh Allen, Stafford, and Mahomes (barely) are less pass-heavy than the overall league. The Eagles, unsurprisingly in 2021, are more pass heavy with lower value as their passing game has struggled.
- Top right – above average pass rates, above average excess points
- Top left – below average passing rates, above average excess points
- Bottom right – above average passing rates, below average excess points
- Bottom left – below average passing rates, below average excess points