Revisiting the analytics-based predictions of the Cowboys 2021 season

The Cowboys’ season came and went, and all the highs of a 12-win season were quickly crushed by the low of losing to the 49ers in Dallas in the Wild Card round. For those who followed along in our weekly analytics roundups, you know that the Cowboys finished the year as the most efficient team in the NFL by DVOA, yet they squandered such a good season with an early playoff exit.

It was a good reminder that analytics, like all other statistics, are merely a measure of what’s come already and not an indicator of the future. The small sample size of an NFL season only adds to the unpredictable nature of football. However, analytics can be a good tool for evaluating upcoming games, and we predicted each regular-season game on the Cowboys’ schedule prior to the season using analytics from previous years. You can find each prediction below:

Weeks 1-2
Weeks 3-5
Weeks 6-9
Weeks 10-11
Weeks 12-13
Weeks 14-16
Weeks 17-18

Revisiting these predictions now that the season is over is an illuminating exercise, but first let’s take a look at how the various efficiency ratings for each team this season correlated to their respective playoff results. Here’s a breakdown of the NFC teams:

As you can see here, the Cowboys and Rams were both top ten in DVOA on offense, defense, and special teams. They were the only teams in the league to do so. The Cowboys got taken down by a 49ers team that was top ten in DVOA on offense and defense and had a higher total efficiency grade than the top-seeded Packers. Meanwhile, the Rams are in the conference championship game and preparing to take on those pesky 49ers.

Of note is that both the Packers and Buccaneers played efficient football most of the year, but their special teams units were both miserable. In the Packers’ case, their league-worst special teams unit played a large role in their loss to the 49ers, who scored their sole touchdown of the game on a blocked punt.

The AFC picture is very, very different. Only three of their seven teams finished among the league’s top ten most efficient groups, and two of them (Buffalo and New England) played each other in the first round; then, Buffalo played the only other top ten team (Kansas City) the next week.

It’s also fascinating to see the Titans, who earned a first-round bye as the top seed, so low in efficiency. With the Titans being such an inefficient team, it’s no surprise they went one-and-done in the playoffs. The Bengals finished with a rare 0.0% DVOA grade, about average in the NFL, but they played two teams that were less efficient in the postseason in the Raiders and Titans to advance to the conference championship game. It’s no surprise the Chiefs are favored by a touchdown in that one.

Alright, having gone through that, let’s check out the regular season predictions. Those analytics-based predictions had the Cowboys finishing 11-6 on the year, which was only one game off of their actual 12-5 record. Here’s a game-by-game comparison:

The first thing to note is that the predictions were spot-on for the first five games of the season. The first deviation was the win over the Patriots, which was predicted as a loss that would be “a very close game for a few reasons, but it’s hard to bet against the Patriots in Foxborough.” The game, of course, was decided in overtime but ultimately went the Cowboys’ way.

The same goes for the Cowboys’ other overtime game this season, in which they lost to the Raiders. Obviously, there was no way that anyone or any model could have predicted all of the turmoil the Raiders endured this season. However, the analytics suggested a solid offense and an improved defense (which is what we saw, for the most part) that would play it close against the Cowboys but ultimately falter. Again, this was a close one but it went the other way in overtime.

Then there were the Vikings and Broncos games, two outcomes that turned out to be just what they felt like at the time: statistical anomalies. Cooper Rush beating the Vikings in Minnesota is unlikely, just as the Broncos curb-stomping the Cowboys in Dallas was. Yet, we live in the universe where both of those things happen, and it’s a great back-to-back reminder of the unusually high level of variance this sport has.

Perhaps the biggest miss from these predictions was expecting the NFC East to be better. Washington’s defense was always likely to regress some, but going from third in defensive DVOA to 27th was a huge surprise; Chase Young and Montez Sweat’s various absences played a part in that. Only getting one game from Ryan Fitzpatrick also played a part in the Football Team’s offense not making a bigger improvement.

As for the Giants, they added a lot of pieces to both sides after a promising close to their first season under Joe Judge. This tidbit from the Week 15 prediction - in which Dallas was projected to lose to New York - ended up being eerily prescient:

But between this defense being so young and the offense featuring so many new pieces, odds are good that they’ll once again be playing their best ball come December. Unless, of course, things fall completely off the rails earlier in the year. But that does seem unlikely at the moment.

That’s exactly what happened, though. Judge fired Jason Garrett during the year and their offense got worse. Then Judge went viral for several post-game press conference rants, which ultimately led to his firing after the season.

Then there’s the Eagles, who took advantage of the fifth-easiest schedule by DVOA and snuck into the playoffs, only to be thoroughly smacked around by the Buccaneers. The fact that they went 0-6 against playoff teams, including two losses to the Cowboys by a combined score of 92-47, reinforces the analytical suggestion that they’d be a bad team in 2021. They just also happened to play a bunch of other bad teams.

All in all, these predictions turned out to be fairly accurate. An overestimation of the divisional foes was a major weakness, while two overtime games and two outliers accounted for the other discrepancies in these picks. Considering how random and lucky football can be, this guessing game didn’t turn out half bad.