-
Bill Connelly, ESPN Staff Writer
In a roundabout way, Texas Tech gamed my SP+ ratings in 2022.
I have long written about a measure I call postgame win expectancy — it takes all the plays in a given game, tosses them into the air and, based on the stats that tend to be most predictive (and which feed into SP+), says, “With all these stats, you could have expected to win this game X% of the time.”
Add up all your postgame win expectancies from a given year, and you get what amounts to a second-order win total: “With all these stats this year, you could have expected to win X games on average.”
In Joey McGuire’s first season in charge, Texas Tech’s second-order win total was 5.8 — not bad against a top-15 schedule that featured five SP+ top-20 opponents, and good enough for 32nd overall in SP+. Their actual win total? Eight. In terms of the difference between wins and second-order wins, it was the third-largest overachievement in the country. Only Sun Belt champion Troy and BYU topped it.
The Red Raiders beat Oklahoma 51-48 despite a 39% postgame win expectancy. They comfortably won games against Kansas (43-28) and Ole Miss (42-25) despite pretty tight stats (postgame win expectancy: 65% and 69%, respectively). And against Texas in September, they pulled off an absolute magic act, beating the Longhorns 37-34 in overtime despite getting outgained by 2.3 yards per play (7.1 to 4.8) and generating a far worse overall success rate (48% to 37%). Postgame win expectancy against Texas: 2%. On average, these four games should have produced about 1.8 wins; Tech won all four.
Was this lucky? Perhaps. Tech did recover all three fumbles in the win over Texas — two of their own (on drives that produced 10 points) and one from Texas in overtime (which ended the game) — and fell on an unsustainable number of loose balls throughout the season. But they also rigged these games in their favor with an extreme willingness to go for it on fourth down.
The Red Raiders were 6-for-8 on fourth downs against Texas, and for the season they went for it a whopping 52 times, easily the most in the country.
Of these 52 attempts, Tech converted 33 and went on to score 129 points after these conversions. After the 19 failures, they allowed 44 points. That’s a profit of plus-85 points, or nearly a touchdown per game over 13 contests. Yes, there was a bit of good fortune in overachieving their win expectancies. But it wasn’t all luck. McGuire played the odds, used fourth down as a win opportunity and won games with it.
Your coach could do the same.
A fourth-down boom
Fourth downs have long been the first front in the football analytics war. Just as baseball managers long gave away too many outs with bunts and stolen base attempts, and basketball teams gave away points by parking members of their supporting cast 16 to 20 feet away from the basket (available for a 2-point mid-range jumper) instead of 23 feet away (where they could take a shot worth 50% more points), football teams long gave away scoring opportunities by punting the ball away or, in some situations, settling for a field goal (“always take the points” and whatnot) on makeable fourth downs.
In their seminal “The Hidden Game of Football” — now available via re-release — authors Bob Carroll, Pete Palmer and John Thorn ran their early iteration of win probability numbers and concluded that “going for it on fourth down happens a lot less often than the calculations warrant. And the cases of trying to get a first down with more than a yard to go are as rare as Tuesday touchdowns.”
Granted, they wrote this in 1988, long before MACtion began giving us Tuesday touchdowns, but as was the case in about 100 different ways with this book, they were on to something. And they were screaming into the void for quite a while.
But in the 2000s, nerds like Football Outsiders’ Aaron Schatz (inspired after finding a used copy of “Hidden Game”) and Advanced Football Analytics’ (and now ESPN’s) Brian Burke joined the chorus. In the 2010s, with help from companies like Championship Analytics (CAI), coaches began listening.
At the college level, the rates of teams going for it on fourth downs were increasing by only a trickle until about 2018, when rates quickly and significantly jumped.
Interestingly, the maximum and minimum go-for-it rates haven’t changed all that much in this period — Hal Mumme’s New Mexico State had a go rate of 51% in 2006, higher than anything we’ve seen since, and Craig Bohl’s Wyoming had just a 6% rate in 2022, lowest in FBS since 2016. But the number of teams hopping aboard the fourth-down train has surged: In 2004, only five teams went for it at least 30% of the time; in 2022, four topped 40% and 33 topped 30%.
What caused the shift? As much as anything, it appears to stem from the success of early adopters. Army earned a healthy number of positive headlines while going 31-for-36 (86%) on fourth downs in 2018 (and going 11-2 while doing it). They attributed a lot of their success to “the book,” aka the weekly binder they received from Championship Analytics, and Ole Miss’ Lane Kiffin has referenced analytics for his willingness to go for it 46% of the time in 2020-21. That certainly didn’t hurt when it came to increasing the customer base for CAI and other companies — CAI alone went from 17 FBS clients in 2006 to 85 this year.
Fourth-down profit
Previously, I mentioned the idea of profit, of comparing the points scored after a fourth-down conversion to the points allowed after a turnover on downs. Let’s revisit that.
I enjoy this concept. Looking at profit allows a coach to attempt different levels of aggression or conservatism — what matters is that they make it pay off. Kiffin’s Rebels, for instance, did not in 2022. Ole Miss went for it 38% of the time last year, eighth most in FBS but a downgrade from previous seasons. Their conversion rate fell from 63% to 49% (which probably contributed to his decreased aggression), and they ranked just 106th in fourth-down profit (0.8 points per game).
Meanwhile, Michigan went for it just 21% of the time (87th) but picked its spots well and profited by 5.3 points per game (ninth). Jim Harbaugh’s Wolverines converted 17 of 21 attempts (81%, second in FBS) and scored 81 points after those 17 conversions. Opponents scored only seven points after the four failures.
In all, only 13 of 131 teams finished with a loss instead of a profit on fourth downs. The simple fact that most fourth-down attempts come far from your own goal line means the points you can expect to allow are never going to be huge. Still, the range between the team with the most profit (West Virginia, with 8.3 points per game) and the least (Colorado, with minus-2.9) was more than 11 points per game. That’s massive.
(* “Normal” = what I’m defining as normal circumstances: 1 to 3 yards to go, less than 60 yards from the opponent’s end zone and in a situation with between 20% and 80% in-game win probability, per ESPN’s FPI. Looking at this range certainly cuts the sample size a bit for particularly good or bad teams — Georgia faced only two such decisions all year because the Bulldogs were winning too much, and Boston College faced only one, in part because of, well, the opposite reason — but it’s an attempt to regulate for some key factors, at least.)
Here’s that same data in chart form:
We see a lot of different approaches here, with a lot of different ranges of success. Colorado went for it more frequently than Maryland in 2022, but Maryland generated nine more points’ worth of “profit” per game, in part because the Terps were good at converting (61%, 27th in FBS) and the Buffaloes very much were not (37%, 117th).
Still, the overall trendline is clear: More fourth-down attempts means more profit. Of the 12 teams that profited by at least five points per game, nine went for it at least 30% of the time, and of the 13 teams whose profit was at zero or fewer points per game, 10 went for it 20% of the time or less.
Profit can have a pretty obvious impact: The teams that profited by at least six points per game here went a combined 21-12 in one-score finishes (a win percentage of 0.636), while teams profiting by at least five went 41-27 (0.603) and teams profiting by at least four went 76-62 (0.551). Everyone under four? 253-260 (0.493). Failing in this aspect doesn’t necessarily hurt you, but succeeding can make all the difference in the world.
It can certainly steal some extra wins, too. Eleven teams profited by more than five points per game on fourth downs, and they overachieved their second-order win totals by a combined 11.2 games (meaning, they won 11.2 more games than SP+ would have expected).
Fourth downs by no means explain everything in the close-games universe — other second-order overachievers such as BYU, Middle Tennessee, NC State and Ohio in no way benefited from fourth-down aggression. But it’s an extra tool for the toolbox. And for at least a few teams, conservatism likely played a role in tamping down win totals.
If aggression can lead to profit, and profit can lead to an extra close-games boost, that certainly seems to encourage coaches to read “the book,” follow the analytics and take a few more swings on fourth down. It certainly worked for Joey McGuire.