NFL teams are taking 4th-down risks more than ever — but still not often enough

The Analytics Hero of the NFL’s opening week very easily could have been a goat. Late in the third quarter on an afternoon full of hope in Cincinnati, the Bengals faced a fourth-and-1 from their 30-yard line, leading 21-7. And Zac Taylor, their 38-year-old head coach, faced a decision that 10 years ago would have been a no-brainer. Any rational coach would have punted. No sensible fan or pundit would have questioned him.

But Taylor coaches in the 2021 NFL, a league that is, slowly but surely, being remodeled by nerds. So he didn’t punt. He went for it. The decision, on top of being “gutsy” and “ballsy,” as his players said, was mathematically sound. This time it backfired, but it mirrored similar decisions that led the Bengals to victory. And it was emblematic of a changing league.

NFL teams went for fourth-down conversions 51 times this past weekend, the highest single-week total in modern league history, up from 37 in Week 1 last year, 26 in 2019 and 22 in 2018. That year, in Week 1, teams went for it on fourth down eight times prior to the fourth quarter. Just three years later, that number ballooned to 33.

The Bengals didn't shy away from going for it on fourth down and, in the end, were rewarded with a victory. (Andy Lyons/Getty Images)
The Bengals didn't shy away from going for it on fourth down and, in the end, were rewarded with a victory. (Andy Lyons/Getty Images)

And while circumstantial factors could explain some of the uptick, advanced statistical models show a significant trend. EdjSports, a data analytics company that works with NFL teams, studies fourth-down decision making and the effect that each choice has on a team’s win probability. It analyzed every relevant and non-obvious fourth-down decision from Week 1 this year and last. It found that, cumulatively, suboptimal decisions on last season’s opening weekend cost teams a cumulative 170 percentage points of win probability.

This past weekend, that cumulative cost was 104 percentage points — almost equal to the win probability gained from fourth-down aggressiveness.

In situations where another model viewed the go-or-kick decision as a “toss-up,” offenses stayed on the field 30% of the time, up from 14% last year.

Analytics have long suggested that NFL coaches, a notoriously risk-averse species, should take more risks on fourth down. Gradually, as PhDs infiltrated football departments, coaches began to listen. The opening week of the 2021 season presented the clearest evidence yet that the nerds’ recommendations are taking hold — and, in some cases, helping win games.

Reward outweighs the risk

Fourth-down aggressiveness doesn’t always snatch headlines. When it does, the headlines often highlight decisive plays or failure. But the best examples are usually buried deeper in games, and they were plentiful this past weekend. Sean Payton’s Saints, for example, went on fourth down twice on the same first-half drive. Math supported both choices. Jameis Winston converted twice, the second time for a touchdown that accelerated New Orleans’ beatdown of Green Bay.

In New York, meanwhile, Vic Fangio’s Broncos converted on fourth-and-7 in the first quarter, drove for a chip-shot field goal, and never looked back.

In Vegas and Kansas City, the Ravens and Browns both jumped out to early leads thanks to analytically savvy decisions. Kevin Stefanski’s Browns passed up field goals on each of their first two drives, converted on fourth-and-3 and fourth-and-1, turned six points into 15, and very nearly upset the Chiefs.

The optimal decision, of course, does not always produce optimal results, and Bengals fans very nearly re-learned that the hard way. Taylor’s choice to go on fourth down from his own 30 increased the Bengals’ win probability by nearly 2 percentage points, according to EdjSports. Their subsequent turnover on downs, obviously, did not. Less than a minute later, the Vikings scored and halved the lead. Eventually, they sent the game to overtime.

Taylor didn’t regret his choice. “I don’t take back any decisions,” he later said. This was the correct one, just like his earlier decision to go on fourth-and-1 in field-goal range. Joe Burrow got that first down, and three plays later, Joe Mixon scored a touchdown.

So Taylor wasn’t fazed when, in overtime, his offense faced yet another fourth-and-1 in their own territory. Mathematical models said it was a “DEFINITE GO.” The Bengals indeed went, and because they did, they won.

The moral of the story is not that aggressiveness guarantees success. Several teams — the Panthers, Jets, Bills, Colts, Bears, Ravens and Raiders — made sound fourth-down decisions and came up short. The statistically supported lesson, instead, is that sound processes yield positive results more often than not. And coaches, finally, seem to be taking that to heart.

Mike McCarthy's decision to kick a field goal late in the the Cowboys' loss to the Buccaneers was deemed the worst decision of Week 1. (Julio Aguilar/Getty Images)
Mike McCarthy's decision to kick a field goal late in the the Cowboys' loss to the Buccaneers was deemed the worst decision of Week 1. (Julio Aguilar/Getty Images)

The worst coaching decision of Week 1

They still, however, have a long way to go. From Week 1 alone, EdjSports unearthed 75 fourth-down decisions that it characterized as “suboptimal.” Many of the 75, in reality, were toss-ups; the difference between kicking and going for it was, in some cases, fractions of a percentage point in win probability. Major errors were far more scarce than in Week 1 of last year, and perhaps more scarce than ever before.

Still, the cumulative cost of those 75 decisions was significant. And the vast majority of the mistakes were conservative ones.

Perhaps the most notable was one that very few seasoned football watchers questioned. On Thursday night, Dak Prescott expertly engineered a late-fourth-quarter drive, and set up Greg Zuerlein for what many assumed would be a game-winning field goal. But on fourth-and-6 from the Tampa Bay 30, with 1:29 remaining, down two points, EdjSports’ model and betting markets alike recommended that the Cowboys keep their offense on the field. Their reasoning, in short, was the likelihood of what ultimately played out.

EdjSports founder Frank Frigo explained in more detail. Even after Zuerlein drilled a 48-yarder — which, of course, was no sure thing — the Cowboys were still more likely to lose the game than win it. Tom Brady with the ball, a timeout, a one-point deficit and 1:24 on the clock was still the betting favorite. By going for it, the Cowboys could have taken the game out of Brady’s hands. A conversion could have made Zuerlein’s field-goal attempt easier and decisive.

EdjSports’ model calculated all of this in real time. Retrospectively, it dubbed head coach Mike McCarthy’s choice to kick the worst coaching decision of Week 1.

There were other maddening ones, too. The Panthers punted from the Jets’ 33-yard line. The Falcons kicked on fourth-and-goal from the 3, and, 55 minutes later, realized it was one of their best and only shots at a touchdown. They finished the game with six points in a blowout loss to the Eagles.

Perhaps the most dangerous decision was Bears head coach Matt Nagy opting to punt on fourth-and-2 with his team in Rams territory. Because it reeked of recency bias, or an inversion of the “gambler’s fallacy.”

The Bears had tried, correctly, to convert two fourth downs in the first quarter of their season opener. They’d failed, and Nagy seemingly let those failures cloud his consideration of the third situation, which models indicated was a clear “go.” That cognitive bias is the only thing that could halt this rising tide of aggressiveness across the NFL.

Coaches have, for decades, made detrimental decisions on fourth down because they were afraid to fail. Whereas successful fourth-down conversions are often attributed to the players who executed them, failures are pinned on coaches and weigh them down.

The worry, then, is that coaches could regress as the season wears on, and that’s why experts will wait, patiently, for the sample size to grow before dubbing 2021 a fourth-down breakthrough. But past seasons suggest that aggressiveness will not wane as the season progresses. And the mathematicians behind the trend certainly won’t recommend fewer risks. They’ll recommend more.