A curious thing has happened in the last few years of the National Football League: teams started going for it on fourth down more. A lot more. In the first week of the most recent NFL season, teams attempted a fourth down conversion 33 times before the fourth quarter, four times as many as the first week of 2018. In October 2020, Ben Baldwin created the 4th down decision bot on Twitter, and in two seasons it has gained 32,100 followers. Fourth down aggressiveness is probably the NFL’s biggest trend, one that’s probably here to stay.
It used to be that, for a variety of reasons, most fourth down decisions were rote and required no real decision making. If it was fourth down on you side of the field, you punted. If you were in field goal range, you kicked a field goal. If you were down by a lot of points at the end of the game, you went for it. NFL coaches are famously risk averse, and doing what everybody else did was the least risky thing possible.
But in the past two or three seasons, analytical data finally filtered down to teams and through the skulls of coaches. That data was consistent: teams should be trying to convert fourth downs way more often and in significantly more field positions and score situations. Data shows that NFL teams convert fourth-and-one plays at a 65% clip, with quarterback sneaks successful nearly 83% of the time. Getting an extra set of downs is just so much more valuable than field position accrued from punting.
Perhaps most interesting in the world of fourth down conversions is that you don’t need to know anything about the stats to intuit that the traditional, conservative, punt-first philosophy is often wrong. Football fans are so used to their coaches making poor decisions that they often get physically uncomfortable when the opposing team goes for it on fourth down in situations where, until the past few years, teams never did. That intuition is right: teams should have been going for it all along, and we all just got used to inefficiencies.
On Wednesday afternoon—April 21—the Kansas City Royals were down 1-0 in the bottom of the ninth inning against the Minnesota Twins. Michael A. Taylor, a very fast runner, worked a leadoff walk. Nicky Lopez then stepped up to the plate and attempted a bunt. The key word here is “attempted.” Lopez popped up for what was potentially the easiest out in the game.
Baseball has been around a lot longer than American football has, and more importantly, baseball analytics have been a part of the game for longer than football analytics have. Sabermetrics was an influence on why football teams are going for it more on fourth down, as it should be.
Thanks to decades of baseball analytical work, we know that one of the few operating truths of the practice is that you should never bunt except for in extreme or very particular circumstances. Bunting is bad.
We know this is true broadly because of run expectancy matrices. While complicated sounding, it’s pretty simple: certain baseball states are more likely to score runs than others. For example, you don’t need a run expectancy matrix to know that a bases loaded, no out situation is more likely to result in runs than a bases empty, two out situation. Thanks to huge amounts of data, we can calculate what the average run expectancy is for these varying states. Courtesy of Tom Tango and Retrosheet, here we have the run expectancy matrix for the average number of runs scored from specific base/out states:
..and we also know the likelihood that a run—any amount at all—will score following each base/out states:
You’ll notice that there are no bunting scenarios in which a sacrifice bunt improves your run expectancy. None whatsoever. Giving up an out to move any combination of runners over with any number of outs left lowers how many runs you can expect to score.
Even if you lower the bar and are solely looking at the chance of scoring at all, there are only two situations where giving up an out to move a runner or runners over could potentially improve your odds of doing so: moving a sole runner from second base to third base increases your odds of scoring, and moving runners from first base and second base to second base and third base increases your odds of scoring.
There are two giant asterisks involved, however. The first is that both aforementioned situations are true only if there are no outs on the board; in other words, sac bunting any combination of runners with one out lowers your chance of scoring.
The other catch is even bigger: bunts are not always successful, as we saw on Wednesday when Lopez popped up his bunt attempt. In 2020—the only year other than this season where no pitchers stepped to the plate—there were 403 plate appearances that ended in a bunt. Of those, 126 were sacrifices and another 133 ended up as singles; 144 were simple outs. In other words, nearly 36 percent of bunt attempts ended up as outs without moving over the runner. Bunts are often couched as automatic. They are not.
But, ultimately, you don’t have to know anything about bunts to know why they should be seldomly used. When a player squares up to bunt, he doesn’t strike fear into the defense or the fans watching the game. There’s no feeling of apprehension or dread, like when you watch an opposing team attempt a crucial fourth down. When a player makes a bunt attempt, they are actively hurting their team’s chance at winning. That’s just how it is. And if we never see another bunt again this season, it would be an improvement to the game.