According to Football Outsiders, over the last three years, 60% of all passes have gone to wide receivers, 21% to tight ends, and 19% to running backs. There are some players who are position hybrids, of course, but as a general rule, wide receiers catch about 56.3% of passes, tight ends have a 63.1% catch rate, and running backs record a reception on 72.4% of their targets. In theory, those numbers should help us figure out which teams (and passers) have completion percentages that are artificially high (or low) because of a high number of passes to running backs (or receivers).
Let’s use the 2013 Chiefs as an example. Last year, 57% of Kansas City passes went to wide receivers, 28% to running backs, and 15% to tight ends. If we use the league-average numbers on passes to players at each position, we would “expect” Kansas City to complete about 61.9% of their passes if the Chiefs were an average passing team. That’s a number that’s slightly higher than league-average rates because the Chiefs threw very often to running backs and not so often to wide receivers.
But as it turns out, projecting a 61.9% completion percentage isn’t the best projection for the Chiefs. I performed a regression analysis to best-fit completion percentage based on the percentage of team pass attempts to running backs and percentage of team pass attempts to tight ends.1 Here’s that formula:
Completion Percentage = 0.525 + 0.135 * RB_%_of_Targets + 0.263 * TE_%_of_Targets
Since the Chiefs threw 28% of their passes to running backs and 15% to tight ends, Kansas City would be projected to complete 60.2% of passes. Now let’s say those numbers were reversed: if 28% of KC attempts were to tight ends and 15% to running backs, the projection would be 61.9%. That’s not an insignificant difference, but the interesting part is that the variables are moving in the opposite of the direction we would expect.
After all, running backs have significantly higher catch rates than tight ends, but this analysis tells us that the more teams throw to tight ends (at the expense of pass attempts to running backs), the higher the expected completion percentage. The formula places a weight on the tight end number that is nearly double the weight placed on the running back variable, when we would normally think the reverse would be true. So what’s up with that?
That’s a good question. Does it mean that throwing passes to tight ends is somehow correlated with quarterback ability, or that throwing passes to running backs is correlated with the lack thereof? It’s easy to think of Drew Brees and Tom Brady throwing lots of passes to the tight end, but Matt Schaub, Sam Bradford, and a rookie Cam Newton also show up on the list of teams with the highest percentage of passes to the tight end. Meanwhile, no single team threw more often to running backs over the past three years than the 2013 Saints. I looked at teams that threw the most passes to tight ends at the expense of their running backs. The ten teams with the strongest preference in that direction had a completion percentage of 60.3%, while the ten teams that were most RB-heavy (at the expense of tight ends) were at 60.1%. That’s pretty weird considering teams *should* have a much higher completion percentage when throwing to backs.
Here’s something else pretty weird. The teams that threw the most to tight ends (instead of running backs) completed 69% of passes to RBs, 57% to WRs, and 63% to TEs. The teams that threw the most to running backs instead of tight ends completed 73% of passes to RBs, 54% to WRs, and 65% to TEs. Again, I’m not sure what the answer is here, so I’ll open it up to the group.
It seems reasonable to suggest that great tight ends open things up for the offense. As a result, passes to tight ends could be a proxy for talent at the position, and that could lead to better numbers for the wide receivers and running backs. So I’ll look to you folks: what do you think is driving these results?
- Because the percentage of passes to RBs, WRs, and TEs would sum to 100% for every team, you wouldn’t run the regression on all three variables. [↩]