Fantasy Football: Regression Candidates Through Week 9

Regression is a big deal in fantasy football.

Who doesn't want to get in low on a player whose peripheral data is better than his fantasy point tallies are? Who wouldn't want to trade away a player who is playing way above expectation -- or perhaps avoid him in daily fantasy formats?

Expected regression is something that we can apply both to season-long fantasy football but also daily fantasy, as well.

So, I went through and paired up various parts of our Net Expected Points (NEP) model to help see where the biggest discrepancies are between players' fantasy points and their expected fantasy points.

Keep in mind that NEP correlates really (really) strongly to half-PPR fantasy points, so a lot of players are playing close to their actual fantasy point output. But it's the outliers that we should focus on.

Quarterbacks

Positive Regression Candidates

Matt Ryan is an interesting outlier at the top of this list because he's been an efficient passer this season but does lose out on touchdowns to Todd Gurley near the goal line; the Atlanta Falcons are 19th in pass rate from inside the 10 this season.

Philip Rivers remains a frustrating quarterback because of how the Indianapolis Colts spread the ball out amongst their pass-catchers, leaving little upside for any specific players. Rivers himself, though, should be better than he has been so far overall.

Joe Burrow is a surprising addition to the positive regression list because he has had big games from a yardage standpoint this year, but the data on him is even better than the fantasy point production. He's the slightest tick above the NFL average in per-drop back efficiency but has significant volume along with it.

Negative Regression Candidates

I'm not going to pump the brakes on Russell Wilson; he has been too good this year based on the advanced data, and the Seattle Seahawks are sixth in pass rate from inside the 10. That helps him pile up touchdowns (14, fourth-most in the league from inside the 10).

What's more notable is players such as Drew Lock and Carson Wentz rating even better than they should've based on their NEP data. Lock's Week 9 performance is the prototypical negative regression example when he threw for 313 yards and 2 touchdowns despite a per-drop back Passing NEP of 0.04 (the NFL average is 0.14).

Running Backs

Positive Regression Candidates

So. No. Don't buy low on Frank Gore. The reason he's here is he gets volume (108 carries) and has no touchdowns. He has as many successful rushes (44) as Alvin Kamara, and successful rushes correlate incredibly strongly to rushing fantasy points. It's vital that we contextualize some of these standouts. Gore is underperforming for an obvious reason: his offense is bad, and he's not particularly efficient either.

Overall, this is a boring list of underperforming running backs. The reason for it is that these are primarily backs with volume but no production tied to it, specifically with touchdowns. These are some of the biggest outliers in terms of yardage gained versus touchdowns scored at the position.

Josh Jacobs and Joe Mixon are the most obvious ones to dig into. They're not that far off their expected paces overall but still should have more production than they do. Jacobs' lack of performance comes from the rushing aspect. He has 73 successful rushes, ranking behind only Dalvin Cook (83) and Derrick Henry (81) but is yet to show the yardage they have (843 for Cook and 855 for Henry to 588 for Jacobs).

Negative Regression Candidates

Similar to Russell Wilson, we don't need to steer clear of Dalvin Cook, but it's important to understand why he's outperforming expectations. Primarily, it's touchdowns. He has, based on his underlying data, 5.3 more touchdowns than expected, most of any running back. In fact, no other back is outperforming expectation by even 4.0 touchdowns. However, the Minnesota Vikings run the ball 71.1% of the time from inside the 10.

The next three on the list -- Melvin Gordon, Aaron Jones, and D'Andre Swift -- are all above their touchdown expectations by at least 1.3 scores, a common thread for the over-performing running backs. It makes sense: little separates running backs in terms of yards per carry, but it's volume, touchdowns, and receiving work that really make a difference.

Wide Receivers

Positive Regression Candidates

Tyler Boyd makes the list because of a lack of touchdowns (he should have roughly 5.1 but has only 3). This is despite a pristine 0.97 Reception NEP per target, a mark that ranks him sixth among 34 receivers with at least 50 targets. Listen to the list he trails: DK Metcalf, Will Fuller, Julio Jones, Davante Adams, and Calvin Ridley. That's nuts. And Boyd has done it on a robust 68-target sample.

Negative Regression Candidates

Adam Thielen (3.7) and Mike Evans (3.8) are two of four receivers who are at least 3.0 touchdowns above expectation (the others are also on this list: Tyreek Hill at 3.8 and Christian Kirk at 3.1).

While DK Metcalf found himself at the bottom of the positive regression list, his teammate, Tyler Lockett finds himself near the top of the negative regression list. Metcalf has put up 18.6 more Reception NEP than Lockett and is significantly better per target (1.15 to 0.83). They have both outperformed touchdown expectations by at least 2.0 (it's 2.1 for Metcalf and 2.6 for Lockett). With the better data on Metcalf, he's the one we should prioritize.

Tight Ends

Positive Regression Candidates

It's wild to see George Kittle and Travis Kelce here, but they are so much better than other tight ends in our expected points data that they still stand out. Kittle makes it mostly because he has had just 2 touchdowns but should have roughly 4.5 at this point, and Kelce isn't that far off his point total but is still shy 0.9 touchdowns from expectation. They're so good.

Negative Regression Candidates

On the flip side, we have Jimmy Graham, who has 5 touchdowns but should have just 2.9. Waller has been about on point with his touchdown totals but just has a lot of garbage in his production -- if you compare it to someone with similar volume.

On 78 targets, Kelce has 753 yards, 6 touchdowns, and 73.6 Reception NEP. On 71 targets, Waller has just 390 yards, 4 touchdowns, and 30.1 Reception NEP. Kelce is nearly twice as efficient per target than Waller is.