Anyone who has spent any time studying football analytics knows one truth: teams are not aggressive enough on fourth down. For example, in situation-neutral contexts, it’s always advisable to go for it on 4th-and-1. The value of possession has become increasingly important in the modern game, where offenses are so adept at gaining yards and scoring points, and the likelihood of conversion is so high that the trade-off of 40-50 yards of field position for a chance to keep possession is almost always worth it. Possession, after all, is worth about 4 points: if having 1st-and-10 at the 50 yard line is worth 2 points, then being on defense in that situation is worth -2, making the swing between having the ball and not having the ball worth 4 points.

So are NFL teams becoming smarter when it comes to 4th down decision making? I looked at all 4th-and-1 plays since 1994 that (i) came in the first three quarters, (ii) with the offense between the 40s, and (iii) with the team on offense either leading or trailing by no more than 10 points. From 1994 to 2004, teams went for it on these 4th-and-1 situations about 28% of the time. Then, from ’05 to 2014, teams went for it 35% of the time. But over the last two years, offenses have stayed on the field for these fourth downs over 40% of the time both years. Take a look:

There have been 495 such 4th-and-1 attempts since 1994, at least according to PFR. On average, those plays have added 1.4 expected points to the offense just by making the decision to go for it. Teams moved the chains 72% of the time in these situations, and ran 84% of the time and passed 16% of the time.

For what it’s worth, the conversion rate has not been trending up or down: it’s held steady around 72% over the last two decades. It was 69% in 2014, 73% in 2015, and 69% last season, so if teams are going for it more frequently, it’s not because it’s easier to convert. Rather, this provides some evidence that the NFL is slowly figuring out that possession is the name of the game. Similarly, the pass/run ratio has been relatively consistent, although there’s large variance due to small sample size issues. But over the last 4 years, the conversion rate is 67% and the pass ratio has been 16%.