In the end, the "flip the switch" theory turned out to be wrong. If anything, the switch was flipped the other direction in Super Bowl LV. The Kansas City Chiefs came out and got their asses kicked, playing their worst game of the year according to DVOA. They had a lower rating than they had for the Week 17 game where they sat their starters, even after significant adjustments for the quality of the Tampa Bay team they were facing.
Kansas City had its lowest offensive rating of the season at -26.1%. (Their only other negative offensive game was -3.3% against Miami when Patrick Mahomes threw three picks). Their defensive rating was better, 12.5% after adjustments, which was in line with a number of their lesser defensive performances from the 2020 regular season. But it was a lot lower than their strong defensive performance in the first two playoff games. The offense, at least, could blame injuries along the offensive line for its dismal performance. The defense just got outplayed and outcoached.
|DVOA (with opponent adjustments)|
|VOA (no opponent adjustments)|
Tampa Bay finished the regular season second in overall DVOA, trailing only their rivals in New Orleans. That rating, of course, is complicated by injuries and the weirdness that was the COVID season of 2020. Take out the game where New Orleans got to play a Denver team with no quarterback, and Tampa Bay would have been No. 1 for the year. But also take out the games where New Orleans had to start a backup quarterback, and they would regain the No. 1 spot. If we include the playoffs, though, it's no contest. Tampa Bay ends up No. 1 for the regular season plus the postseason at 34.8%. New Orleans, even crediting their defense for the Kendall Hinton game, is No. 2 at 28.8%. Green Bay is third, and Kansas City is fourth.
Tampa Bay's final DVOA including the playoffs ranks 12th among Super Bowl champions since 1985, with an asterisk. The asterisk comes because I moved to a new version of DVOA in 2020 and have not yet had an opportunity to go back and re-run all the playoffs back 35 years, so this is comparing Tampa Bay in DVOA v7.3 to other teams in DVOA v7.0. They are 22nd among all teams since 1985 if we combine the regular season and the postseason. One of the teams that is higher than the 2020 Buccaneers is the 2002 Buccaneers at 41.0%.
Tampa Bay finished the regular season third in weighted DVOA, which lowers the strength of older games. They were behind Buffalo and Green Bay in that metric. They finish the postseason No. 1 in weighted DVOA as well, narrowly over Green Bay 39.44% to 39.40%.
I wrote this on Twitter, but I will write it again here, as a bit of a mea culpa. Football Outsiders took a lot of criticism this year about where Tampa Bay and Kansas City ranked in DVOA. We took criticism for having Tampa Bay too high -- they were in the top three every week since Week 5 -- and for having Kansas City too low. This was not the normal criticism from fans who yell at us to just watch the film because we're wrong about their favorite team. There was a lot of particular criticism from the rest of the analytics community about anybody who didn't have Kansas City No. 1 in their particular metric or power ranking. It was criticism from people whose work I appreciate and value. And I will admit, I really started to doubt my own work and my own metrics. I worried that I had to be missing something if so many smart people were in agreement that DVOA was completely wrong. After Kansas City's big playoff wins, I was convinced that everyone else was right and I had to be wrong. I tried to come up with some sort of adjustment to account for the fact that Kansas City was the first NFL team to "flip the switch" after coasting through the regular season. In hindsight, it was clearly a mistake.
In the end, it turned out that DVOA was seeing something. It saw that Tampa Bay was better than its 11-5 record, in particular that it was a very good team even when it was "struggling" at 7-5 with a number of close losses to good opponents. It saw that Kansas City's string of close wins were emblematic of a team that was very good but not by any means historically dominant. And it turned out that Kansas City's two strong playoff games weren't any kind of "flipping the switch." They were just a very good team having two excellent games, which happens all the time.
Not that DVOA was perfect, of course. The same people who were critical of Football Outsiders for having Kansas City too low were also critical of Football Outsiders and anyone else who had New Orleans rated highly, and that complaint turned out to be accurate. And we shouldn't pretend that DVOA in any way was predicting some kind of Tampa Bay blowout in this game. With the adjustments we tried to make to account for Kansas City "turning it on" in the playoffs, the FO+ picks still called for Tampa Bay to cover the spread. But without those adjustments, the projected line would have been Tampa Bay by only 0.3 points.
Obviously, there were some extenuating circumstances to Kansas City's loss. We will never know how Kansas City might have played if they had their offensive line healthy, or if they had not shuffled their linemen due to injuries. Of course, Kansas City played the second half of the season, including Week 12's win over Tampa Bay, with a lot of those same linemen (albeit in different positions) and they did not look anywhere near as bad as they did last night. Should we have expected the Kansas City offensive line injuries to have some impact? Yes. Enough to make the game a 31-9 blowout? I don't think so. There was also Patrick Mahomes' turf toe, although given how much he scrambled around behind the line of scrimmage and the accuracy of his passes that bounced off his receivers' face masks, I don't think the turf toe was really hampering Mahomes too much.
Nonetheless, despite these circumstances, I think this game is a data point in favor of DVOA and, more old school, a data point in favor of point differential as a predictive tool. Someone suggested to me on Twitter that it doesn't make sense to learn a lesson from one game, which is just small sample size. But of course, the opposite is true. The small sample size isn't Tampa Bay beating Kansas City; the small sample size was Kansas City allegedly being the first team to make point differential not matter because they could "flip the switch." We have 35 years of ratings showing that teams with higher DVOA are more likely to beat teams with lower DVOA. We have even more years of point differentials showing us that the Pythagorean projection is telling us something about team quality.
As far as ratings and stats, I think the biggest story was not that Tampa Bay was better than Kansas City during the regular season, but that this was a year with a lot of good teams and no dominant team. Despite all the people, including analytics people, who insisted that Kansas City was "inevitable," no team was inevitable. In that way, it was the perfect year for a wild-card team to win the Super Bowl. It was the perfect year for coaching to matter -- not in-game decision making, which we talk about around here a lot, but game planning and the decisions Tampa Bay's coaches made to play against type in ways that flustered the Chiefs and helped win a championship.
Thanks again to everyone for a great 18th season at Football Outsiders. We'll start our offseason coverage soon along with announcing the winners of the Football Outsiders reader awards next week. Publication of 1983 and 1984 DVOA is coming sometime in February or early March.
Normally, we publish DVOA and Quick Reads for the Super Bowl together in one post. Today we did them separately. You'll find Quick Reads with individual Super Bowl stats here.