NFL Fourth Down Decisions and When Going For It Pays
Fourth and two at the opponent's 38, six minutes left, down four. The punt team trots on, the ball lands at the 12, and the broadcast cuts to a graphic saying the decision cost two percent of win probability.

Fourth and two at the opponent's 38, six minutes left, down four. The punt team trots on, the ball lands at the 12, and the broadcast cuts to a graphic saying the decision cost two percent of win probability. Half the living room yells at the television and the other half says analytics ruined football. Both reactions skip the part that actually matters, which is what the coach was weighing.
I have watched this argument get louder every season since the win probability models went mainstream, and most of the heat comes from treating fourth down as a question with one right answer. It is not. It is a question with a right answer that moves depending on distance, field position, time, score, and how good your own team is at gaining two yards. That last variable is the one nobody puts on the screen.
The arithmetic underneath the call
Start with conversion rates, because everything else is built on them. Pooled across recent seasons, fourth and one converts somewhere around 70 percent of the time. Fourth and two sits closer to 58. By fourth and four you are near a coin flip, and past fourth and seven you are under 40 percent and falling.
Then the alternative. A punt from your own 40 nets about 40 yards after returns and touchbacks, so the opponent starts near their own 20. A punt from the opponent's 38 nets far less, usually 20 to 24 yards of actual field position, because the ball either goes in the end zone or gets fair caught inside the 15. That shrinking return is why the models get aggressive as you cross midfield. You are risking a possession to gain very little.
| Situation | Rough conversion rate | What the punt or kick buys | Model lean |
|---|---|---|---|
| 4th and 1, own 35 | 70 percent | Opponent starts near their 20 | Go, by a small margin |
| 4th and 2, opponent 38 | 58 percent | About 22 net yards, no field goal | Go, clearly |
| 4th and 4, opponent 25 | 48 percent | 43 yard field goal, around 80 percent | Kick, usually |
| 4th and 8, own 30 | 38 percent | Opponent starts near their 25 | Punt, comfortably |
Those lean columns are the output of a win probability model. The two best known public ones are the EdjSports framework and Ben Baldwin's fourth down calculator built on nflfastR play by play data. Both do the same basic thing: simulate the rest of the game from each branch, average the outcomes, and report the difference in your chance of winning.
The trend in behaviour is real and large. A decade and a half ago teams attempted fourth down roughly 450 times across a season. Recent seasons have run past 700, and the increase is concentrated exactly where the models said it should be, between the 40 yard lines with two or fewer yards to gain.
Where the model is strong and where it gets soft
A league-average conversion rate is a useful starting point and a bad finishing point. Some teams convert fourth and one at close to 80 percent because they have a quarterback who will run it and two tight ends who can seal an edge. Some convert at 55 because their short yardage package is a handoff into a stacked front and everyone in the stadium knows it.
Baldwin's calculator lets you nudge the team strength inputs, and when you do, calls that looked obvious start flipping. A bottom-five run blocking unit changes the break even distance by most of a yard, which is enormous in this context. The same goes the other way: injuries to interior linemen quietly remove the aggressive option from a coach's menu, something we go into in our piece on how offensive line injuries decide playoff fates.
The models are also weakest on the things that are hardest to measure. Wind over about 15 miles per hour changes both the punt and the field goal branch. A backup quarterback changes the conversion branch and the comeback branch at once. A kicker who has missed from 48 twice in a month is not the league-average kicker in the simulation, and the coach knows something the data set does not.
Why always go for it is as lazy as always punt
This is the part that gets me unpopular with the spreadsheet crowd. A large share of the fourth down calls that cause arguments on television are worth one to two percent of win probability. That is a real edge and it is also inside the range where two reasonable models disagree with each other.
The calls that actually matter are much bigger and much rarer. Fourth and one anywhere between the 40s is typically worth four to seven percent. Fourth down while trailing in the fourth quarter, when a punt means handing away the ball and the clock together, can be worth 10 to 15. Those are the ones where the gap between the coach and the calculator is not a rounding error. If a team fixed only its fourth quarter decisions and stayed conservative everywhere else, it would capture most of the available gain.
Meanwhile the first quarter fourth and one from your own 30 is where the models are most aggressive and least convincing. The edge there is small, the variance is enormous, and the model is assuming a neutral game state that a 14 point first quarter hole will destroy. I am not telling you the model is wrong. I am telling you that spending your credibility with a locker room on a one percent call in the first quarter makes it harder to make the seven percent call in the fourth.
Which brings up the thing that explains coaching behaviour better than any curve. A failed fourth down is a visible, replayable, nameable mistake. A punt that quietly lowers your win probability is invisible. Coaches get fired for the first kind. Until that changes, expect the gap between optimal and actual to persist, and expect it to narrow fastest among coaches with contract security.
How to read it live, before the play clock runs out
Watch the personnel on third down. If an offense sends out a heavy package on third and three rather than spreading out, the staff has already decided it is playing two downs. The substitution pattern tells you the plan before the coach signals anything.
Keep the kicker's range line in your head. For most teams that is the opponent's 37 or 38, which is a 55 yard attempt. Inside that line, the field goal branch exists and the models get more conservative. Outside it, the real choice is go or punt, and punting from the 42 buys almost nothing.
Note the clock thresholds. Inside about six minutes, possessions become scarce and the value of keeping the ball rises sharply. A punt with four minutes left while trailing is usually a worse decision than the same punt with twelve minutes left, even from the identical spot.
And watch who is under centre on the sneak. The quarterback sneak from the one or two yard line converts at a rate most coaches still underrate, somewhere north of 80 percent behind a decent line, and it is the single most efficient play in the sport. Teams that have figured this out run it from fourth and one in situations where the broadcast is still showing a punt graphic. The rest of our NFL coverage goes through more of these rule and strategy wrinkles, including how these decisions compound into seeding, which we unpack in the playoff tiebreakers.
FAQ
What is the break even conversion rate on fourth down?
There is no single number because it depends on field position. Deep in your own territory you might need to believe you will convert 65 percent of the time to justify going. Near midfield the bar drops into the 40s, and inside the opponent's 40 on fourth and short it can fall under 40 percent, because the punt gains you so little.
Does the two point conversion follow the same logic?
Same framework, different inputs. Two point tries convert around 48 percent league wide and the extra point is close to 94, so going for two is roughly break even in raw points and becomes correct once the score difference makes one point less useful than two. The decision tree is published and most staffs carry a laminated version of it.
Which teams are most aggressive?
It shifts year to year with staff turnover, but the pattern has been consistent: teams with young head coaches hired from analytics friendly front offices, and teams with mobile quarterbacks who make short yardage cheap. Aggressiveness rankings published each season track how often a team made the model-preferred call rather than how often it went for it.
Has all this aggression actually helped?
Modestly, and not evenly. League scoring and fourth down attempts both rose over the same period, which makes clean attribution hard. The clearest finding is that the teams gaining the most are the ones adjusting for their own personnel rather than following a chart, which is the least satisfying answer and probably the true one.
More in NFL
NFL
NFL Trade Value Chart and How Teams Price Draft Picks
I have been talking myself out of the same bad trade in Madden since about 2009.
NFL
NFL Fifth-Year Option Explained and How Teams Use It
Every year in the first days of May, a batch of headlines lands saying a team has declined a fifth-year option, and the replies...
NFL
NFL Compensatory Picks and How Teams Earn Them
Every March a list comes out that hands some teams four extra draft picks and others none, and the explanation from most...