Pats, Texans among overrated, underrated in NFL power ranks
Chiefs top Power Rankings as only unbeaten team
FPI's message to the ESPN Power Rankings panel is straight out of the Aaron Rodgers playbook: R-E-L-A-X.
Although there are plenty of teams the two systems agree on, when they differ, it is by and large because the voters are making much more aggressive moves off of this season's action than FPI is comfortable with.
Sixteen games is a small sample. Four is nothing. Sure, we've learned things in those games, but why throw out the vast quantity of information we had going into the season? That's not a good way to go about predicting what's coming next.
Below are five teams FPI thinks are significantly underrated or overrated by ESPN's post-Week 4 Power Rankings.
Power rank: 7
FPI rank: 2
The voters really want to bet against Tom Brady? Against Bill Belichick? After what, two losses?
FPI thinks that's nuts.
How long do we have to watch the Patriots be on top of the world before we forgive a loss here or there? Apparently, 14 straight years of double-digit wins isn't enough.
Look, we know what this is about: the defense.
In four games this season, the defense has added negative expected points every time. It has clearly cost New England so far, and no one would argue that point.
But there's also this: Defense is less consistent than offense. Past defensive performances don't predict future defensive performances to the degree that offensive performances predict future offense. In the Patriots' case, that variance works in the favor of their future prospects.
That being said, FPI agrees that the defense will be a liability. Going forward it thinks New England's defense is 1.3 points worse than average per game. That's ninth worst in the league (and anyone who thinks the Patriots' defense will be in the same league going forward as those of the Colts or Browns ought to watch a few more Colts or Browns games).
So a good offense plus a bad defense equals a mediocre team, right? Not exactly. We all know there's a little more nuance to it than that.
How about we try this: An exceptionally strong offense plus a weak defense equals ... the second-best team in football. New England's league-best offense, by FPI's standards, is 8.6 points above average per game, and 3.4 points better than that of the New Orleans Saints, the second-best offense. That makes up for the defensive issues and then some.
The Patriots will be just fine. No, much better than fine.
Power rank: 30
FPI rank: 17
Here's what FPI cares about: performance against expectation, while taking into account circumstantial factors such as opponent strength and game location. It has a separate expectation for the offense's strength depending on which quarterback is expected to be under center, and it keeps in mind the information it held on teams in the preseason.
Here's what it doesn't care about: past wins and losses. Those just aren't the best way to predict football, and that's what we're doing here, forecasting what is to come. We already know what just happened.
So consider the case of the 0-4 Giants. Would they have dropped two spots in the power rankings if Bucs kicker Nick Folk's field goal at the end of the game had been off target and Big Blue had won the game? We didn't learn a whole heck of a lot about the Giants in the moment, and that's why FPI doesn't put much stock into it.
Instead, the model considered the circumstances of the game -- that the Giants were on the road facing a better team and lost by only two -- and actually moved them up by a hair after that one.
The Giants have likely put themselves in too deep a hole to reach the postseason (just a 2 percent chance, FPI says), but they aren't one of the worst teams in the NFL.
Power rank: 10
FPI rank: 18
So far, our numbers are buying Deshaun Watson. He's the NFL leader in Total QBR and clearly has helped the Texans en route to their 2-2 record.
The model does account for individual quarterbacks, and it has made significant adjustments for Watson already. It thinks the Houston offense will operate at a higher level with Watson at quarterback now than it did in Week 1. But FPI isn't ready to anoint him yet, and with good reason. It's been just four games for him, too - three-and-a-half, actually -- and that isn't enough time to figure out whether he truly is as good as he has been.
Also keep in mind that a tremendous amount of Watson's value came from one play -- the 49-yard scramble for a touchdown in Week 2 for which he deserved and received a tremendous amount of credit. Watson's raw, unadjusted QBR for the season is 81.9. Take away that one play and it would drop to 70.1. Still good -- third best, in fact -- but it shows how incredibly valuable that one play is. The good news is FPI is smart enough to cap the impact of extreme plays such as that one when considering information for its predictions.
And remember, Watson isn't alone out there. While the offense was magnificent in the Texans' most recent game, a 57-14 win over the Titans, that hasn't been the case in every game. In two of Houston's four contests this year, the offense posted negative expected points totals. So combining all that information with the prior expectations -- the model didn't think much of the Texans' offense coming in to the year -- FPI believes the offense is a couple of points below average.
Granted, FPI believes in the defense. But it's just not enough to make us think Houston is better than average at this point.
Power rank: 17
FPI rank: 7
There is no singular reason FPI thinks so much more highly of Minnesota than the voters do.
The difference quite possibly derives from a combination of some "jack-of-all-trades syndrome" and the lack of star power at quarterback.
FPI considers the Vikings above average, but not exceptional, at all three phases of the game. That isn't the kind of team structure that leaps off the page, so perhaps they get overlooked as a result.
Additionally, quarterbacks like Sam Bradford and Case Keenum don't inspire much confidence. Frankly, FPI doesn't think much of them, either, but it believes in the offense as a whole, which it projects to be slightly above average going forward. The model does build in short-term uncertainty about which QB will start for the Vikings, though it doesn't see a tremendous gap between Keenum and Bradford.
Interestingly, although FPI believes the Vikings are more than a point better than the Lions, Minnesota's chances of reaching the postseason (48 percent) and winning the NFC North (22 percent) are worse than those of Detroit (and Green Bay), thanks in large part to the fact that Minnesota is simply a game behind those two division rivals.
Power rank: 16
FPI rank: 24
This discrepancy is all about Jared Goff.
The top pick in 2016 has been one of the surprising stories of the 2017 season so far, as he has played well so far.
But FPI is still scarred from last year, when the Rams' offense was absolutely dreadful when Goff was at quarterback. The model thinks more highly of the L.A. offense with Goff at quarterback now than it did, but it isn't about to forget last year and we wouldn't want it to: When there are so few data points, we have to hang on to every one of them.
The young QB has a Total QBR of 60.4 this season (10th best) and if Sean McVay really has helped turn Goff (and the entire offense) around, then FPI would indeed be underestimating them. But we ought to find out how much of a change there has been in Los Angeles over the next four or five games.
To this point, the Rams have faced the Colts, Redskins, 49ers and Cowboys. The Redskins have a solid defense, but the other three rank in the bottom 12 in FPI.
Next up are the Seahawks, Jaguars, Cardinals, Giants and Texans: All five rank in the top 12.
Additionally, although Aaron Donald earns plenty of headlines, the model thinks this defense is a little below average, too.
All in all, FPI thinks this is a below-average football team for the rest of the season -- though it does have a 27 percent chance to reach the postseason thanks to its 3-1 start.
Hank Gargiulo contributed to this story.
For more from ESPN Analytics, visit the ESPN Analytics Index.