It certainly looked like the Buffalo Bills could be the best team in the NFL at the midway point of this season. They destroyed the Miami Dolphins and Houston Texans by a combined score of 75-0. They beat the Kansas City Chiefs handily. They lost to the Tennessee Titans, but only by a field goal. They looked very strong.
Then the Bills went out in Week 9 and lost to the Jacksonville Jaguars, a team that is currently 2-10. That game looked terrible, but the following week the Bills had another dominating win, 45-17 over the Jets. Then the Colts ran all over them in a dominating 41-15 victory. The Bills looked down again until they came back and dominated the Saints 31-6 the next week.
The constant roller coaster from week-to-week has us wondering: Who are these Buffalo Bills, anyway? Are they the really great team that easily beat the Chiefs? Or the mediocre team that couldn't even beat the Jaguars? The answer is that they are both. Overall, the Bills are pretty good, but more importantly they are inconsistent.
In fact, the Bills may be the most inconsistent team in the NFL since at least 1983.
Football Outsiders measures inconsistency using a stat called variance (explained here). Instead of looking at raw scores, the DVOA ratings break down every play and compare success to a league average adjusted for situation and opponent. These ratings filter out some of the effects of luck and can tell that a 9-7 loss to Tampa Bay was probably a good game while a 9-7 loss to Jacksonville was definitely not.
Variance looks at the DVOA for each individual game and determines just how consistent or inconsistent a team has been that season. And with a current variance of 44.5%, the 2021 Bills come out as the most inconsistent team we've ever tracked in stats that go back to 1983. For comparison, the median variance for an NFL team over a 16-game season is 14.6%.
Below we'll look at the five most inconsistent teams as well as the five most inconsistent teams that made the playoffs and the most inconsistent Super Bowl champion since 1983. After all, the Bills could definitely find themselves with that title by the end of this season.
Jump to:
Most inconsistent teams
Most inconsistent playoff teams
Most inconsistent Super Bowl team
Most consistent team
Most inconsistent teams overall since 1983
5. 2003 San Francisco 49ers (7-9)
Variance: 35.0%
This was the first year of Dennis Erickson's two seasons as San Francisco head coach, and the last year for Jeff Garcia as the 49ers' starting quarterback. The 49ers opened the season with a 49-7 stomping of the Chicago Bears. Three weeks later, they were on the other side of a similar score as they lost 35-7 to Minnesota.
At midseason, the 49ers had big wins 24-7 over Tampa Bay and 30-10 over a St. Louis Rams team that finished 12-4. In between, they lost 16-13 in overtime to the worst team in football by DVOA, the 4-12 Cardinals. Later that season, San Francisco had consecutive weeks where they lost 44-6 to the Baltimore Ravens and then beat the Cardinals in a rematch 50-14.
Incidentally, San Francisco's variance in 2003 has nothing to do with three games where backup Tim Rattay had to start in Garcia's place. The 49ers went 2-1 in those games, including the aforementioned Rams win.
4. 1989 Pittsburgh Steelers (9-7)
Variance: 35.4%
The 1989 Steelers were led by legendary coach Chuck Noll and not-so-legendary quarterback Bubby Brister. They significantly over performed their underlying stats. The Steelers made the playoffs as a wild card despite finishing 20th in DVOA and getting outscored during the season 326-265.
A lot of the variance for the 1989 Steelers comes from one loss which happens to be the worst single game in the history of the DVOA metric. The Steelers opened the season with a 51-0 loss to the Cleveland Browns which rates out at -159% DVOA. The Steelers had 53 total net yards with five first downs and lost the turnover battle 8-0. It was one of three shutouts for the Steelers. They also lost 27-0 to the Houston Oilers and 20-0 to the Chicago Bears, and in between those two shutouts, the eventual AFC champion Denver Broncos beat them 34-7.
Yet this same team beat the Detroit Lions in Week 4, 23-3, forcing four turnovers. They beat the Browns in the rematch, 17-7. They beat Dan Marino and the Miami Dolphins by the score of 34-14 and then shut out the New York Jets by the score of 13-0. Overall, the 1989 Steelers somehow won more games than they lost despite having some absolutely horrible losses. They even beat the Oilers in the playoffs before losing to Denver in the divisional round.
3. 2007 Denver Broncos (7-9)
Variance: 35.9%
This was a fairly unremarkable late-Shanahan team with Jay Cutler at quarterback in his second season, notable mostly for the number of blowouts they played both winning and losing. The Broncos lost 41-3 and 23-3 to the San Diego Chargers. They lost 44-7 to a Detroit Lions team that finished the year 29th in the league in DVOA.
But they also beat the Chiefs 41-7 in Week 14. They also have a bit of a hidden blowout, a 34-20 Monday night victory over Tennessee in Week 11 where the Broncos gained 8.0 yards per play (not counting kneel downs to end the game), picked off the Titans twice and forced four fumbles (although they only recovered one of them).
You would also expect an average team like the Broncos to beat bad teams and lose to good teams. But the Broncos played better than expected against good teams (a 31-28 win over Pittsburgh, sixth in DVOA that year) and they played worse than expected against bad teams: in two games against Oakland, which ranked 30th in DVOA, they had an overtime win and a 34-20 loss.
2. 2005 San Francisco 49ers (4-12)
Variance: 38.2%
In general, the most inconsistent teams of all time mixed big wins with bad losses. The 2005 49ers are special because they mixed narrow wins with horrendous, ridiculously bad losses. The 2005 49ers are not only currently the most inconsistent team in DVOA history. They're also currently the worst team in DVOA history, even though they won four games!
That's because the 49ers' losses were abysmal. They lost 28-3 to Indianapolis. They lost 52-17 to Washington. They lost 41-3 to Seattle. At least those were all good teams. They also lost 42-3 to Philadelphia -- still defending NFC champions, but a team that finished 6-10. Then they also won four games, but only by a combined score of 15 points.
The 2005 49ers had two of the 10 worst single games since 1983. They had four games with single-game DVOA below minus-100%. This season, there have only been five games with single-game DVOA that low -- by every team in the NFL combined.
1. 2021 Buffalo Bills (7-4)
Variance: 44.5% (through Week 12)
The Bills have huge wins over bad teams: 35-0 over Miami, 40-0 over Houston and 45-17 over the New York Jets. They also have big wins over good teams: 38-20 over Kansas City, currently ninth in DVOA and climbing and last week's 31-6 victory over New Orleans, 12th in DVOA and fifth on defense.
The Bills then combine those wins with a curb stomping by Indianapolis, 41-15, and two losses to bad teams, Pittsburgh and Jacksonville. So Buffalo's DVOA ratings have been all over the place this season. Buffalo has five games with single-game DVOA over 60%. New England has three, and no other team has more than two. But no team in the DVOA top 10 has more games with below-average DVOA.
Variance should get lower with a larger sample size, so there's a good chance Buffalo's variance at the end of the year won't be quite this high. But even if we only compare them to past teams through Week 12 instead of past full-season teams, the Bills still come out with the highest variance since 1983.
Most inconsistent playoff teams overall since 1983
5. 1998 New York Jets (12-4)
Variance: 31.5%
The last thing Bill Parcells would want is an inconsistent team, but this was a very inconsistent team. The 1998 Jets had nine different games with DVOA of 50% or more, the most of any team that season. However, they also had three very bad games with DVOA below -50%, all losses to poor opponents.
In Week 2, they lost 24-10 to a Baltimore Ravens team that finished 6-10. In Week 6, they lost 30-10 to a St. Louis Rams team that finished 4-12. And in Week 11, they lost 24-23 to an Indianapolis Colts team that finished 3-13. Other than an overtime loss in Week 1 to San Francisco, those were the Jets' only losses of the regular season. The Jets eventually lost in the AFC Championship game to the eventual Super Bowl champion Denver Broncos.
4. 2013 Philadelphia Eagles (10-6)
Variance: 31.6%
This Eagles team wasn't necessarily inconsistent in Chip Kelly's first season as head coach. They had a high variance because their 3-5 start featured some particularly bad performances while their 7-1 finish featured some particularly good ones.
The worst day was in Week 4, when the Denver Broncos beat the Eagles 52-20. At least that was a really good Broncos team that finished 13-3.The Eagles also lost at midseason to two mediocre NFC East opponents in games that were not as close as the final score: a 17-3 loss to Dallas and a 15-7 loss to the Giants. The very next week, the Eagles turned it around to wallop the Raiders 49-20.
Their second half of the season also featured a 54-11 victory over the Chicago Bears. The Eagles had 10 wins but they also had three very bad games with DVOA below -60%, even worse than the 1998 Jets. Philadelphia won the NFC East but fell to New Orleans in the wild-card round.
3. 1997 Tampa Bay Buccaneers (10-6)
Variance: 33.2%
This was Tony Dungy's second season in Tampa Bay and his first season with a winning record. Tampa had some strong wins, such as 13-6 over the 49ers (who finished 13-3) and 27-7 over the Patriots (who finished 10-6).
But like the 2013 Eagles, the 1997 Bucs had three games with DVOA below -60%. One of those games was actually a win, 19-18 over the Arizona Cardinals in Week 5. Arizona finished 4-12 and 27th in DVOA, yet they outgained Tampa Bay 364 yards to 167 yards in this game. Tampa Bay won thanks in part to a blocked punt return for a touchdown and two missed field goals from Arizona's Kevin Butler, including one as time expired.
Tampa's other really bad games in 1997 included a 27-9 loss to Detroit in Week 7 and a horrendous 31-0 shellacking by the New York Jets in Week 16 where the Bucs only managed 111 yards of offense. The Bucs beat Detroit in the wild-card round and then lost to Green Bay in the divisional round.
2. 1987 Indianapolis Colts (7-5)
Variance: 33.6%
There's a little bit of an asterisk here because we're not considering the three strikebreaker games Weeks 4 through 6, only the 12 games played by the "real" Indianapolis Colts.
But those games bounced all over the place. For example, the Colts beat Miami 40-21 in Week 10. They followed that up with a 24-0 shutout loss to New England in Week 11. One week later, they beat the Oilers 51-27. Then they beat the Browns 9-7. That's a quality win since Cleveland finished the season third in DVOA, but you'll notice that in four weeks their total points went 40, 0, 51 and 9. A week later, they lost to Buffalo 27-3 with only 130 yards of total offense, just two weeks after the 51-point outburst against Houston. The Colts eventually lost a rematch with Cleveland 38-21 in the divisional round.
1. 1989 Pittsburgh Steelers (9-7)
Variance: 35.4%
One of the top-5 most inconsistent teams since 1983 is also the most inconsistent playoff team. Mainly for the same reasons we discussed earlier.
Most inconsistent team to win the Super Bowl since 1983
1996 Green Bay Packers (13-3)
Variance: 25.2%
Wait a minute, you're probably saying, "How can an inconsistent team go 13-3 and win the Super Bowl?"
Too often, we confuse "inconsistent" with "mediocre." But really bad teams can be inconsistent if, like the 2005 49ers, they mix small victories with huge losses. And really good teams can be inconsistent if, like the 1996 Packers, they mix small losses with huge victories.
The Packers really dominated the league in 1996, especially at the start of the season, winning their first three games by an average of 29.7 points. Then they went out in Week 4 and lost 30-21 to a Minnesota Vikings squad that finished the year 21st overall in DVOA. Green Bay finished the year with six games that had DVOA over 60% and ten games that had DVOA over 40%.
But they also had some close calls against bad teams, such as a 13-7 win over the Buccaneers (who finished 6-10, 22nd in DVOA). They barely outgained a very bad Rams team (6-10, 28th in DVOA), although the final score of 24-9 was more impressive than the play-by-play. Two weeks later they beat the Broncos 41-3, although that wasn't quite the Broncos team that went 13-3 that year (they were sitting John Elway). Eventually, the Packers marched to the Super Bowl easily, winning their three playoff games by a combined score of 100-48.
The median variance for Super Bowl champions is 14.0%. So, yes, Super Bowl winners are generally a little more consistent than your average NFL team.
However, last year's champions the Buccaneers had the highest variance of any team in 2020 at 21.4%. They weren't as high in variance as the 1996 Packers because they didn't have as many dominant victories, but they mixed impressive wins with huge losses (38-3 to New Orleans in Week 9) and narrow escapes over bad opponents (25-23 over the Giants in Week 8 or 31-27 over Atlanta in Week 15).
Most consistent team since 1983
2016 New York Giants (11-5)
Variance: 2.4%
The 2016 New York Giants, the most consistent team in DVOA history, played 11 one-score games -- including nine out of their first 10 contests. Their biggest win and biggest loss were both by two touchdowns.
That biggest win was a 27-13 victory over the Cleveland Browns (1-15) which came out close to average with DVOA's opponent adjustments. Their biggest loss, 24-10 to the Minnesota Vikings (8-8), wasn't that bad since the Giants averaged more yards per play than their opponent. Each of the Giants' games fell between minus-15% and 40% in single-game DVOA. So far this season, every team except Minnesota has at least two games that fall outside that narrow span.
The 2021 Vikings currently have 2.9% variance and would be one of our five most consistent teams ever if the season ended today. All of their games -- except one -- fall between minus-15% and 20% DVOA.
**** Technical explanation of variance: Variance measures how far each number in a data set is from the mean. So in this case, we take the squared difference between a team's single-game DVOA and mean DVOA for the season. Average out those squared differences, and we get variance.