≡ Menu

Chip Kelly, Bill Belichick and Riverboat Ron (née Rivera) are the three coaches most associated with aggressive fourth down decisions. In week 15, all three faced a key fourth down decision, and each situation provided a good teaching moment.

Philadelphia goes for it on 4th and 1 on their own 24, down by 15, 3rd quarter, 6:26 remaining

This was an unconventional decision, but pretty clearly the right one. Here’s what Kelly said after the game:

“I thought we could’ve made it, and I also thought if we don’t make it we’re in trouble,” Kelly said. “If we can’t get half-a-yard, maybe it tells you what the day’s all about. But you’ve gotta think at 4th-and-half-a-yard we can get half-a-yard. They didn’t blitz. It wasn’t like there was an all-out coming at us. We need to come off the ball and get some movement at the point of attack and dig ourselves out of that hole right there. We hadn’t gotten anything going at that point in time so hoping we could jump-start something there.”

The bold decision did not pay off when LeSean McCoy was stopped on 4th-and-inches, and that failure enabled the leaders of the conservative moment to begin crowing. Going for it on 4th down is no guarantee, but from 2010 to 2013, runs on 4th-and-1 have converted 68.6% of the time (excluding runs in the red zone). And one would think a dominant rushing team like Philadelphia would be in much better shape. Philadelphia leads the league in both rushing yards and yards per carry, while the Vikings have a below-average run defense. According to Advanced NFL Stats, Philadelphia should have gone for it if the team had a 50% chance of conversion, and the Eagles probably had a 75% chance of converting here.

But the frustrating part of the analysis is that the anti-stats movement claimed that it was “too early” for Philadelphia to go for it. As a general rule, it is never too early to go for it on 4th-and-inches, and that applies even more strongly when talking about the Eagles offense.
The phrase “too early” has a very specific meaning. It implies that there are stages in a game where a team should make a decision that — while not being a good idea normally — makes sense because it increases the team’s variance. For example, no one suggests that teams should onside kick after every score, because there is a lower expected return from onside kicks than from regular kicks. But when trailing late in the game, teams are willing to trade a lower expected return for the higher variance associated with onside kicks.

The flip side of this can be seen in kneel downs. Obviously taking a knee has a very low expected return, but it makes sense when teams are looking to minimize variance. When someone claims that it is “too early” to do something, what they are saying is “we are not at such a late stage in the game that a team should consider trading expected return for higher variance.”

But going for it on 4th-and-inches (or even 4th-and-1) does not bring a lower return. In fact, going for it in this situation brings a higher return, on average, than punting. How do we know this? Consider the three outcomes:

1) Your opponent gets 1st-and-10 from (roughly) the 24. This is worth +3.8 expected points to the other team.

2) You have 1st-and-10 from (let’s be conservative) the 25. This is worth +0.6 expected points to your team.

3) Your opponent has 1st-and-10, following a punt, from (on average) the 36-yard line. This is worth +1.2 expected points to your opponent.

So punting puts your team in a situation worth -1.2 points. If we assume a 68.6% conversion rate, then going for it will provide, on average, -0.8 points.  It’s true that going for it increases the variance associated with the event, but that doesn’t make it “too early” to go for it. This was a slam dunk decision for Kelly, even though the outcome failed.

Side note: One argument often proferred to explain why you “can’t go for it” in this situation is that it will demoralize the team if you fail. Following the failed fourth down conversion, (a) the Eagles forced a three-and-out, resulting in a field goal for Blair Walsh; (b) the Philadelphia offense went 70 yards for a touchdown, (c) the Eagles defense forced an interception, and (d) Philadelphia’s offense scored another touchdown.

Carolina goes for it on 4th-and-2, with the ball on the Jets 14, up by 10 in the 3rd quarter, 7:55 remaining

Aggressive play calls in these situations are what have inspired the Riverboat Ron moniker. And going for it was the right call according to ANS: going for it is worth 2.7 points, on average, while kicking is worth just 2.3 points. There are more specific factors that point in the direction of going for it, like Carolina being an excellent short-yardage team and the 10-point lead.

The difference between a 13-point lead and a 17-point lead is pretty significant. Up by 13, the Panthers are vulnerable to two Jets touchdowns, while a 17-point cusher is much safer. On the other hand, using a win probability model, the break-even success rate where going for it is the right decision jumps from 50% (using expected points) to 70%. Reasonable people could differ about this call; one could argue that this is the flip side to the “too early to go for it” reasoning.  In this case, perhaps it was too late to go for it, as Carolina may have been better off trading some expected return in exchange for a lower variance.

But the reason this decision makes the list is because of what happened after the play: the Jets stopped the Cam Newton pass intended for Steve Smith, and promptly drove down the field and scored a touchdown. Now trailing by just 3 points, many were pointing to Rivera’s fourth down decision as a turning point in the game. Rivera’s “crazy” decision gave a dead Jets team life, and it may have cost his team the game!

That whole line of bunk was rendered moot by the Jets offense and a Geno Smith pick six, but this type of analysis pops up from time to time. There is no such thing as momentum gained after stopping a team on 4th down; it’s bad enough when people rip a decision based solely on outcome, and not process, but it’s even worse to blame a 4th down call for what happens three minutes later. Right or wrong, there was nothing outrageous about Rivera’s decision, and it was probably the right call. Even right calls can backfire.

New England sends in the kicker on 4th-and-2 from the Dolphins 5, down by 7, 4th quarter, 7:49 remaining

Belichick is famous for unconventional decisions, but he went by the book here. The problem was the book contains bad information.  We’ll get into the math in a moment, but the concept to keep in mind here is that avoiding going for it on 4th-and-short often forces a team to go for it on 4th-and-long later.  That turned out to be the case here, as the Patriots had to go for it on 4th-and-8 with 52 seconds left (converted by Danny Amendola) and on 4th down with 7 seconds left from the Miami 14-yard line.

According to Brian Burke, kicking the field goal lowered New England’s probability of winning from 28% to 23%; that’s based on the Patriots having a 39% chance of winning had they converted, a 14% chance had they failed, and a 55% success rate. Of course, for a team with Tom Brady, putting the game in the hands of the quarterback is an even more attractive option.

After kicking the field goal, New England stopped Miami, and then responded with a touchdown drive to take the lead. But Ryan Tannehill then led a 60-yard touchdown drive to give Miami the lead, and the Patriots were left needing to go 80 yards in 75 seconds. Brady put forth a valiant effort, but it would have been a whole lot easier had the team only needed a field goal.

No one is criticizing Belichick today for sending in his kicker on this 4th down, and that’s precisely why coaches are risk averse. Fail to go for it on 4th down, and few care; go for it and fail, and the sharks begin to circle. But Belichick erred by not going for it on 4th-and-2 in the middle of the fourth quarter, which left the team needed to convert on fourth down from the 14-yard line in the final seconds.

  • Alex

    While I think Chip Kelly probably made the right decision to go for it, I have to quibble with your definition of what “too early” means and your argument against its application to this decision. I would define “too early” to mean “At this stage of the game, the returns on available low variance options are sufficient that the team in question should reject high variance options.” The difference is that your definition and subsequent analysis assumes that it can never be “too early” if the high variance option also has a greater expected return. But I see no reason for this assumption.

    For instance, suppose you believe that the true difference in ability between the Eagles and Vikings is such that the Eagles should outscore the Vikings by 16+ points in 21 minutes. Then it is worth sacrificing 0.4 expected points for the (relative) surety of not losing an additional 2.6 expected points since that extra 0.4 points wouldn’t really change the expected outcome while an unlucky break on the high variance play might. Now suppose that the Vikings are still up by 15 at some later point when you no longer expect the Eagles to make up (or come close to making up) the gap between the teams. In the same situation, it is worth risking the loss of 2.6 expected points for the chance at gaining 1.8 expected points since the expected return of -1.2 points on the low variance option is insufficient to push the expected outcome in the Eagles’ favor. The fact that the high variance option also has a higher expected return is then just a bonus.

    Of course, if you had expected the Eagles to outscore the Vikings in the time remaining, you would have had to have the Eagles as 48+ point favorites going into the game, so on the surface, Kelly’s decision to take the higher variance option looks pretty good, but I don’t see how a 0.4 gain in expected points, by itself, invalidates concerns that it’s “too early” or makes it an obvious decision.

    • Chase Stuart

      That’s a good point; perhaps I should have limited my point to games where the two teams are of even strength. I guess when the Broncos play the Texans, they should take the low-variance options even when they bring about a lower EV. But I still think, in general (and in this specific case), the idea of “too early” makes no sense.

      • Alex

        I wouldn’t say the idea of “too early” makes no sense in general. Take two evenly matched teams in a tie game in the first quarter, and consider two scenarios from this article: the fourth down decision faced by Chip Kelly and the one faced by Ron Rivera. I would argue that punting is the correct decision in the former because even though the expected return of going for it may be higher, the upside of succeeding is still quite low and not worth the risk of failing. If you wait a while to be aggressive, you will likely encounter another opportunity to go for a high variance play with greater upside later in the game. On the other hand, if you are down near your opponent’s goal line, the upside is tremendous and the downside not so bad. In that case, it makes sense to go for it early in a game because you will probably never have a better opportunity to swing the game in your favor with a single play. But I would agree that “too early” doesn’t make sense in most of the contexts where we hear it used in broadcasts.