Last week, I wrote about how the Buffalo Bills and their opponents basically traded off three and outs throughout the season and came out 14 points ahead. In the following days, the article received a ton of great comments and discussion (something that’s very cool about the Buffalo Rumblings community; thanks for having me). I’m not a huge commenter myself, but I read each one and appreciate the thoughtful feedback.
With that in mind, a few comments addressed issues converting third downs into first downs. The Bills had the third lowest first down rate on first and second down plays (21.8%) in the NFL, so third downs were critical to keeping drives alive.
But maybe the low first down rates were a result of the short, but reliable, plays designed in an effort to give the Bills "manageable" third downs. That plan worked, in a way, because the Bills did have fifth-lowest yards to go on third downs. The one problem was that the Bills had a pretty bad first down conversion rate on third down (35.7%, compared to a league-wide average of 38.8% in 2013). Buffalo’s disparity between average yards to go and first downs was very unlike the rest of the league.
The Bills’ issues didn’t come on short yardage situations. They were above the league average conversion rate, especially on rushing plays, with two or fewer yards to go (63.5% on all plays, 73.3% on run plays). Despite having the third highest run rate on all plays in the season, however, Buffalo only ran the ball tenth-most with one or two yards to go on third down.
The reason for that disconnect is that the Bills' run play percentage drops from 74% with one yard to go to 25% with two yards to go. Here’s how the Bills did on third down by distance. Notice the drop-off in rushing play percentage while the difference in success rates don’t diverge too much.
That drop-off in rushing plays is a league-wide occurrence (thanks to Jon Ramsey for pointing that out). In the past ten seasons, all NFL teams have dropped their rushing play call percentage from 73.5% on third and one to 37.6% on third and two. That seems pretty inefficient, given run plays have been more successful than pass plays with four or fewer yards to go. The graph below shows the NFL first down conversion rates and the expected Bills conversion rates (based on logistic regression of the 2013 rates) for rushing and passing plays.
Based on this information, the 2013 Bills would have been expected (based on the regressions) to be above average only with one yard to go. So not only should they (and the rest of the NFL) have run more often on third downs with fewer than four yards to go, but they also need to improve.
In order to figure out how they should improve (and improve the third down play calling analysis) we’d need to chart a few variables on each third down with five or fewer yards to go. There were 114 of those plays in 2013, so those results will be in next week’s post.
Maybe we can expand our research to more plays if we crowdsource the project? Anyone interested? It could be a pretty cool project that would lead to some really awesome analysis. Let me know in the comments!