One of the more-discussed areas with analytics in football is around fourth-down gambles. And the models there generally have a lot of information to work off of, as there have been plenty of cases of fourth downs from most distances over the years, and plenty of cases of teams punting, kicking field goals, or going for it.
But a challenge those models can run into is with unusual combinations of down and distance, score, and game clock that haven’t come up a lot over the years. And some of that happened on the ESPN broadcast of the LSU Tigers-Missouri Tigers broadcast Saturday. There, down 42-39 with 1:15 left, Missouri faced a fourth-and-32 situation on their own 28-yard-line, with all three timeouts left. And coach Eli Drinkwitz opted to go for it rather than punt, much to the consternation of announcers Bob Wischusen and Robert Griffin III and the ESPN Analytics department:
"You've got to play the percentages in football, and that would mean punt the ball." – Robert Griffin III, as a "4th & 10 or more: No-Go – ESPN Analytics" banner is displayed on the screen. pic.twitter.com/5yzaMJ06jt
— Awful Announcing (@awfulannouncing) October 7, 2023
A few key comments there, in addition to the on-screen “Fourth and 10 or more: No Go – ESPN Analytics” graphic seen above:
“Missouri is going to go for this, and that’s surprising.” – Wischusen
“Conventional wisdom would tell you punt the ball back to LSU, try to stop them, use all three of your timeouts to get the ball back. But in this situation, they might feel like going for it on fourth and 32, maybe they don’t get it, maybe there’s a chance they can hold LSU to a field goal, and they might not be as aggressive if they’re already on their own side of the field. I would say punt it away, play defense, and you’ve got to be able to trust…I know Jaylen Daniels has been carving up their defense this entire second half, but you’ve got to play the percentages in football, and that would mean punt the ball.” – Griffin
“Even given an opportunity, Missouri will stay with their offense on the field, on fourth down and 32.” – Wischusen
“There’s just not many plays for fourth and 32, let alone at this point, at this stage in the game.” – Griffin
This then led to further discussion after the play failed, including Wischusen claiming “It has to be a two percent chance or less” (of converting a 4th and 32, without showing those numbers), and suggesting there was “better than a two percent chance” of success from the punt strategy (again, without showing those numbers). That then goes into Griffin saying “I ran the numbers, beep, boop, boop, boop, boop, and yeah, it’s higher to punt the ball away and try to have the defense get the ball back for you” (again, without showing any specific numbers).
"I ran the numbers, beep, boop, boop, boop, boop, and yeah, it's higher to punt the ball away and try to have the defense get the ball back for you." – Robert Griffin III pic.twitter.com/hrymi9J1oD
— Awful Announcing (@awfulannouncing) October 7, 2023
That play didn’t lead to a successful fourth-down conversion for Mizzou. But it did get them around 21 yards, meaning that LSU took over on Missouri’s 49. That wasn’t in field goal range. And while LSU was closer to the Mizzou end zone than they likely would have been after a punt, a first down in either situation would have ended the game in their favor given the clock. So the fourth-down gamble actually wound up having some merits from many perspectives, even though it didn’t work out.
If this play had led to a stop at the line of scrimmage or in the backfield, it would have been worse. That could have let LSU kick a field goal without gaining a first down and take a six-point lead. But even that is maybe only marginally worse in this specific situation; giving up points is bad, but the time with which Mizzou would have gotten the ball back on a full three runs and late-clock field goal would have been limited. That might have necessitated a Hail Mary attempt for a touchdown rather than an attempt to try to get into field goal range and get out of bounds with no timeouts left to set up the FG.
In the end, this led to the Mizzou defense getting a stop, and to them getting the ball back on their own five. That promptly led to an interception for a touchdown, and to LSU’s ultimate 49-39 win. And yes, perhaps Mizzou would have done better than the interception if they’d started another 10 to 20 yards down the field after a punt, a defensive stop, and a punt, but that’s far from clear (and LSU fifth-year punter Jay Bramblett may have been able to get it to the five even if he had started further back; he had a long punt of 57 yards in this game, and has one of 65 on the year).
But from a logical perspective alone, Mizzou didn’t seem terribly hurt by taking a chance on retaining possession with only a field goal needed to tie on a fourth-and-32 gamble (where they only came up 10 yards short) that worked out like a short punt. Perhaps there are analytical numbers that indicate differently, and illustrate that this gamble was a poorer choice than a punt. But the ESPN broadcast certainly didn’t show that, instead just referencing “No Go” from ESPN Analytics (a unit which has come under fire many times before in terms of its conclusions and how they’re presented by the rest of ESPN) and then having Griffin and Wischusen tell viewers how they “ran the numbers” without discussing what those numbers were and how they were calculated. (But yet, there was time for Griffin’s calculator noises.)
The other thing to keep in mind here is what was mentioned off the top: a fourth-and-32 situation obviously has many less past data points to consider than, say, a fourth-and-one situation. And that sample size is something that’s critical to keep in mind with any discussion of crucial-down-gambles. And those all take further dimensions still when it comes to late-game situations where clock, timeouts, and more factor into it (and further reduce the similar sample size). So this wasn’t necessarily the best moment for the broadcast and ESPN Analytics to weigh in so decisively. Especially without showing or talking about the actual percentages involved in those calculations.
[H/T to CJ Fogler on Twitter]
