Can NY Jets gain edge in 2023 with fourth-down decision-making?

Robert Saleh and the New York Jets should rethink their fourth-down approach

Hall of Fame head coach Tony Dungy made a crucial observation that is often misapplied by coaches league-wide.

Dungy used to tell his players, “More games are lost than are won.”

Now, in Dungy’s era of coaching, that was meant to exhort players against making the game-changing mistake. In a league of ever-increasing parity, there are more one-score games than ever before. One error can be the difference between victory and defeat.

For the New York Jets and the rest of the 2023 NFL, though, that phrase must be understood differently. With fewer and fewer possessions due to philosophical changes by offenses intended to minimize turnovers, there are so few chances for each team to score.

How do you win a game in the NFL? By finding a way not to cede any extra opportunity to put points on the board.

This means ditching some of the old-school maxims and aphorisms of the NFL of yesteryear, including “You are what your record says you are” and “Run to set up the pass.” It’s about finding that edge, even if it’s not a popular call. Calculated risk is different from reckless daring.

In no area is this more visible than fourth-down decision-making.

Fourth down philosophy

Until the last five years or so, attempting a fourth-down conversion was generally saved either for a desperate situation or for the most daring of coaches. In fact, from 1999 until 2017, the rate of fourth-down conversion attempts barely moved; it was at 11.8% across the NFL at the turn of the century and at 12.7% in 2017, a fairly small increase.

In 2018, there was a still-small but noticeable uptick to 14.9%, which further went up to 16.3% in 2019. It went higher still to 19.1% in 2020 and hit a height of 20.7% in 2021. This past season, a year in which offense and point-scoring were down across the board, the fourth-down conversion attempt rate was still the third-highest in the last 24 years at 18.9%.

This shift occurred primarily due to a belated turn to analytics across NFL teams. Baseball’s own more sweeping shift toward analytics occurred over a decade ago, and there were isolated teams using numbers to gain an edge for years prior. Although NFL teams have turned to the data for various aspects of team-building for a while, it’s only recently that this same philosophy has been applied to in-game decision-making.

There are many fourth-down decision models out there, some free, some paid, and some not publicly available. Every model attempts to isolate the win probability based on each option—go, kick, or punt. The most popular free version is that of ESPN’s Ben Baldwin, whose fourth-down Twitter bot provides in-game analysis for each decision in real time. (Alas, this bot may be on its last legs with new Twitter price scales for bots, which are untenable.)

Rarely do models strongly disagree about a decision. They are all trained on years of in-game NFL data, and the conclusions drawn about the advisability of a particular fourth-down decision will vary only slightly, even if the provided win probability numbers for each given decision may differ.

One thing every model agrees on, though: NFL coaches should be going for it a lot more often than they do, and in situations that the old-time fan might think are outrageous.

Super Bowl LVII

A classic example to illustrate this occurred in the recent Super Bowl matchup between the Chiefs and Eagles. Nick Sirianni, the Eagles’ coach, ranked fourth in the NFL in fourth-down aggressiveness despite having led in almost all his games. Those fourth-down decisions did not come from necessity, but from a strategy to win the game.

Indeed, the Chiefs dealt with this in the Super Bowl on multiple occasions. Sirianni would send out his sneak team and either get the first down that way or force an offsides penalty. It helped Philadelphia sustain drives and dominate the time of possession.

However, down by one point in the fourth quarter, Sirianni allowed his fear of defying convention to overcome his aggressiveness. The Eagles had 4th and 3 from their own 32. In most NFL parlance, with 10:33 remaining in the game, this is a no-doubt punt situation.

However, Baldwin’s model clearly felt otherwise, as did Football Outsiders’.

—> KC (28) @ PHI (27) <—
PHI has 4th & 3 at the PHI 32

Recommendation (STRONG): 👉 Go for it (+4.6 WP)
Actual play: 🏈🦵 A.Siposs punts 38 yards to KC 30, Center-R.Lovato. K.Toney pushed ob at PHI 5 for 65 yards (A.Siposs pic.twitter.com/qV1LJFnw6L

— 4th down decision bot (@ben_bot_baldwin) February 13, 2023

By our model, the choice to punt was an 8% win probability loss. And that's before knowing about the return.

— Bryan Knowles (@BryKno) February 13, 2023

This was not a situation of second-guessing; the model recommended going for it before Kadarius Toney nearly returned the punt to the house. Most models do not take punter strength or punt coverage into account. Rather, they are looking at the game situation and the most likely results based on years of in-game data.

In this case, by analytical standards, Sirianni should have gone for it—and he whiffed.

It’s not that this was the decision that lost the Super Bowl. However, Kansas City had not been able to stop Philadelphia in short-yardage situations, and the Eagles’ defense had not been able to stop Patrick Mahomes in the second half. This was their chance to keep pace in the back-and-forth game. Instead, Sirianni punted, setting up Toney’s 65-yard punt return and Skyy Moore’s walk-in touchdown.

Yes, the Eagles tied the game and forced some more Mahomes heroics. Still, that was a moment to swing the momentum back in Philadelphia’s favor after Kansas City overcame a 10-point halftime deficit to take the lead.

Sirianni did not regret his decision following the game.

Nick Sirianni claims 32 out of 32 NFL head coaches would have punted on fourth-and-3 in the fourth quarter, as he did to set up Kadarius Toney's long punt return. https://t.co/IdueJmL5tF

— ProFootballTalk (@ProFootballTalk) February 16, 2023

Reading between the lines, though, what Sirianni is saying is one of two things.

• It did not occur to him to go for it because this is a situation in which it would not occur to any NFL coach.
• It occurred to him to go for it, but he was afraid of looking like a fool if that decision cost the Eagles.

This is where analytics has not yet made enough of a dent across the league. There is much more willingness to go for it on 4th and 1, and on the plus side of the field, teams will be far more aggressive. In their own territory, especially inside their own 40? Teams still rarely do it.

In fact, teams went for it on fourth down from inside their own 40 (without faking a punt) just 6.7% of the time in 2022, 112 total times. The median time remaining in the game was 4:55, and the median score differential was -9 (meaning that the team with the ball was down by nine points). Even then, the median number of yards to go was just three.

In other words, teams did not go for it inside their own 40 unless they faked a punt (which happened only 17 times all year) or they were completely desperate.

How much win probability did they cede by doing so?

He’s right that all 32 would punt. And all 32 would be wrong. I could also have all 32 head coaches take my game theory midterm next week & all 32 would get every question wrong.
Ignore the other coaches & follow the math. It’ll give you an advantage. (We’ll get there one day.) https://t.co/j9khfTMjqp

— Deniz Selman (@denizselman33) February 18, 2023

We have multiple models all saying the punt was a bad decision, all "soft" factors also point to punting being bad (Chiefs offense good! Eagles punt unit bad!), and the actual result of the punt, (and the game for the Eagles) was bad.

And people are still defending the decision!

— Computer Cowboy (@benbbaldwin) February 17, 2023

Sirianni found out the hard way.

2022 Jets

Robert Saleh went for it 19.3% of the time on fourth down in 2022 (not including plays on which there was a penalty or no play), which was roughly in line with league trends. It’s worth noting, though, that he only went for it with a lead once the entire season (in Week 5 against Miami, when the Jets were up 19-17 in the third quarter; the bot gave him a 7.15% win probability boost and a strong “go for it” recommendation).

19 out of Saleh’s 24 outright “go” calls were when the Jets were trailing by double digits. This does nothing to negate the perception of Saleh as a conservative coach, on the whole. He simply had many situations in which the Jets were desperate, and he had no choice but to go for it.

His most significant straight “go for it” decision was against the Vikings when the Jets were trailing 17-3 at the two-minute warning of the first half. The Jets faced a fourth-and-two from their own 43, and a lot of Jets Twitter hated the decision to go, particularly when Mike White‘s pass attempt to Tyler Conklin fell incomplete.

However, Baldwin’s model supported the decision by Saleh.

—> NYJ (3) @ MIN (17) <—
NYJ has 4th & 2 at the NYJ 43

Recommendation (STRONG): 👉 Go for it (+1.3 WP)
Actual play: 👉 (Shotgun) M.White pass incomplete short right to T.Conklin (C.Sullivan). pic.twitter.com/oG29A6oXez

— 4th down decision bot (@ben_bot_baldwin) December 4, 2022

Old-timers and football purists will balk at this. However, doing things the way they’ve always been done just because they’re the way they’ve always been done is foolhardy. Math isn’t everything (says the former math teacher), but increasingly, it’s a big part of the way NFL games are won and lost.

Here are all the times that Saleh went against a STRONG model recommendation in 2022.

The only one of these that is puzzling is the field goal recommendation with the Jets down 22-10 against New England. A conversion would have still left them trailing by two scores with 8:54 remaining. The call to go for it is seemingly correct.

In the rest of the cases, though, it was a case of doing what most coaches would have done—but not necessarily giving the team the best chance to win. This is the kind of thing that a team at a disadvantage must do to steal victory from the jaws of defeat.

Future of fourth down strategy

To take it a step further, NFL fourth down strategy must evolve past relying on the safe play. Maybe teams do not need to go for it every time there is a medium recommendation to do so; after all, there is still something to be said for having a feel for the game and knowing the situation, plus taking into account matchups (which the models do not).

However, when a team is an underdog or playing a high-powered offense, there is no excuse to give the other team any extra possessions. With that number at ever more of a premium due to short passing, high completion percentages, and fewer penalties, teams must jealously hoard their time with the ball and refuse to give it up if there is any doubt.

Baldwin, who is ESPN’s chief data scientist, the creator of most ESPNAnalytics metrics, and the co-founder of the RBSDM NFL database and nflfastR, puts it in even starker terms when referring to the Eagles’ punt decision:

(No team should ever punt on 4th & 3 in a game they are losing but NFL coaches aren't ready for that conversation yet)

— Computer Cowboy (@benbbaldwin) February 17, 2023

The coaches who are first to embrace this concept will get ahead of the curve on what is most important in the league. It’s not surprising that John Harbaugh dealt with criticism for a fourth down “go” call earlier in the season; after all, the Ravens have been one of the best-run franchises in the NFL since their inception.

The key point here is that this kind of fourth-down philosophy is always critical, but never more so than when facing a superior opponent or a superstar quarterback. Don’t punt the ball to Patrick Mahomes unless you absolutely must.

Jets 2023 approach

It’s understandable that Saleh wants to call a conservative game with a Jets team that is predicated on defense and the run game. However, I would argue that this fourth down strategy dovetails with a time of possession approach perfectly: its whole purpose is to keep the ball out of the opponent’s hands, which is the same reason that teams try to establish the run.

Additionally, year-over-year defensive metrics tend to be a lot more volatile than offensive ones. The Jets’ defense is unlikely to be as good in 2023 as it was in 2022. If they come out of the gate expecting that and adjusting their strategy to meet the reality, they will give themselves a better chance to win.

The Jets do face Patrick Mahomes next season, as well as Jalen Hurts, Josh Allen twice, Justin Herbert, the high-powered Miami offense twice, and Dak Prescott. They do not want to give these teams extra life. Keep the sword at their neck all game.

Now, if the Jets do acquire Aaron Rodgers, perhaps the calculus changes. Maybe other teams are afraid of giving the Jets too many possessions. However, I’d argue that the point gets even stronger: what are you afraid of? The chances of converting are that much stronger if you trust the guy at the helm.

It is unlikely that the Jets will actually do this. Conservatism does not suddenly bleed out of a coach’s veins, although maybe Saleh (and Dan Shamash, the Jets’ in-game decision-making coach) will once again read Michael Nania’s tweets and get some ideas.

Ironically, the best coaches in the business, Bill Belichick and Andy Reid, are also conservative in their fourth-down play-calling. Belichick was more aggressive before it became fashionable and then veered dramatically in the opposite direction. Reid’s biggest flaw as a head coach has always been in-game clock management.

Still, these two can get away with it because of their tactical advantage and brilliant game-planning. It would behoove the Jets’ coaching staff to introspect and realize that they do not have this. Therefore, they must be strong in the finer points of the game.

Aggressive fourth-down play-calling is the wave of the future. The question is if the Jets can recognize it early enough to take advantage.