Two of the biggest analytical findings over the past few years in the NFL have been how passing is more valuable than rushing and how play-action is an incredibly valuable form of passing. But while we know passing is more important, efficient, ad valuable, we also know the correct pass-run ratio isn’t 100-0. The league average still hovers around 60-40 pass and teams have not regularly hit 70 percent passing over a full season. So while rushing should be a smaller part of the game, it will still be a part of it. The question then should be how can teams get better at running the ball, since they’re going to do it anyway. There must be a way for a run play to take advantage of the shift towards the pass and use the deception of play-action.
That answer has been around for quite some time: the draw play.
Like how play-action is set to look like a run before the quarterback pulls the ball back to pass, the draw is supposed to look like a pass as the quarterback holds the ball before a delayed handoff to the back. The main reason behind the draw was initially to slow down aggressive pass rushers, but it can work in a similar way to play-action to manipulate linebackers and other second-level players. Instead of pulling them in, it can push them back and open up room to run.
Despite this potentially being a perfect changeup for the modern game, running back draws only accounted for about 2.2 percent of all rushing attempts in 2018 and the number of draws run has dropped in recent years.
This also despite the running back draw being more successful than a standard running play. Last season, teams averaged 0.09 Expected Points Added per attempt on running back draws with a positive play rate (the percentage of plays that produced positive EPA) of 44 percent, per Sports Info Solutions charting. All run plays in 2018 averaged minus-0.03 EPA per attempt with a positive play rate of 43 percent.
That EPA gap is significant, though the overall success rate is not as high. Much of that, though, can be explained by the high failure rate of draw attempts from under center (37.4 average positive play rate). Still, most draw attempts came from shotgun (80 percent) and the positive play rate there (46 percent) was much higher than the standard run play.
The 0.12 EPA per attempt bump is also the exact same as passes without play-action (0.00) to those with play-action (0.12) in 2018.
No team averaged one draw per game last season. Only 14 teams used a draw at least 10 times total during the 2018 regular season. The Los Angeles Chargers were the best high usage team with 5.1 EPA on 14 attempts (0.36 EPA per attempt).
The draw also certainly helps when a back with receiving ability is in the game. For example, the Philadelphia Eagles were one of the most efficient teams on draws (0.63 EPA per attempt on nine attempts) and they had most of their success with Darren Sproles as the back. The New Orleans Saints only ran three draw plays all season, but the best was a 17-yard gain against the Los Angeles Rams with Alvin Kamara. The Cleveland Browns did their best, including a 23-yard gain on 2nd and 16, with Duke Johnson. A good receiving back isn’t needed to feel the impact of a draw, and most of these players are still going to have more production through the air, but it can give a different look for a defense when expecting the pass.
Death to the 3rd and long draw
One circumstance when the draw shouldn’t be used is on 3rd and long. A draw on 3rd and long is an offense admitting they have no desire to pick up a first down or many yards at all, really. There were 31 running back draws on 3rd and 8 or longer last season and only one gained a first down. The league average conversion rate on 3rd and 8 or longer wasn’t great (24.3 percent) but it presented much better odds than the white flag of a draw.
Overall, those 31 attempts totaled minus-5.4 EPA on the ground (minus-0.17 EPA per attempt). Passing in those situations averaged 0.01 EPA per attempt.
The Arizona Cardinals were the worst offender of this with five draws on third and long in 2018 and averaged minus-0.3 EPA per attempt. Though in their defense, the offense was so bad last season a third and long draw was the better option. On pass plays in those situations, the Cardinals averaged minus-0.46 EPA per attempt.
We can also lump in halftime draws in this category, when instead of kneeling to run the clock to end a half, teams run a draw for little to no reason. The halftime draw actually had positive EPA because yards were gained in those scenarios (even further proof the draw can be an effective play when the pass is expected), but it did not have more positive EPA than 3rd and long attempts had negative, so if we take those two categories away, the draw comes out as an even more efficient strategy.
The QB Draw
To this point, we’ve only talked about running back draws but the QB draw can be one of the most effective weapons in a playbook. It doesn’t even need an overly mobile quarterback to take advantage of it. Last year there were 35 designed quarterback draws. Those runs averaged 0.37 EPA per attempt.
Buffalo’s Josh Allen led the league in QB draws with six and he was able to average 0.35 EPA per attempt with a positive play rate of 50 percent. Cam Newton had five with a 0.64 EPA per attempt average and positive play rate of 60 percent. Lamar Jackson didn’t do as well on draws — only one of his four produced positive EPA — but his one success was enough to still average 0.33 EPA per attempt.
There were also a number of quarterbacks who ran only once, but with success. Taylor Heinicke, Newton’s backup, ran once for 1.5 EPA. Case Keenum had one attempt worth 1.8 EPA. Aaron Rodgers had one for 1.9.
Admittedly, we’re working with a lot of small samples here but that’s also the point. NFL teams still have to run, so why not use a play that was significantly more productive than a standard run a year ago? The draw play isn’t going to be the run game’s play-action in terms of volume, where 25 percent of the league’s passing attempts have a play fake (and it probably should be higher). But the draw can use the same type of deception to fool defenses and should also be used on more than 2.5 percent of rushing attempts going forward. Even if an increase in use drops efficiency, there’s still a long fall before the draw becomes as inefficient as a typical run anyway.