What’s Sticky? Statistics You Can Rely On for Fantasy Football

The FootClan
Love the show? Join our community!
Join the FootClan

If you’ve played fantasy football for multiple years, you’re probably familiar with the concept of ‘sticky stats’, or football metrics that tend to persist over time. For example, targets are commonly considered ‘sticky’ because they are consistent across years. On the other hand, touchdowns scored are most often considered ‘not sticky’ because they vary so much year to year; a player might find the end zone a couple of times (or fall just short) solely based on luck.

It’s important, though, to really quantify what we mean by sticky. This will give us an objective measure of what statistics are useful for forecasting performance across years, how confident we can be in making these forecasts, and if our commonly held assumptions need to be re-examined. All data, unless otherwise specified, is from nflfastR.


We’re going to work through the major fantasy positions, as well as the major statistics at each position, and measure the predictive impact across years. One position-statistic example might be WR receptions. We collect seasonal data from all WRs since 1999, filtering out those who missed significant time (7+ games) or didn’t have significant involvement (less than 40 receptions, etc.). For each WR in a given year, we note the reception total in the previous year for that player. Finally, we exclude players who changed teams between years, since this introduces many confounders – new teammates, new scheme – that make it hard to isolate the relationship that we want.

The final result is a dataset for hundreds of WRs that provides their reception total in the current year and reception total in the previous year. We then run a regression to predict current year yardage using previous year yardage; this is called an autoregressive model, as we are using the lag (previous year data) to predict the current year’s data.

Ultimately, things boil down to one number: the coefficient in the model tells us how many additional receptions we predict in the current year for every reception in the previous year. For example, if the coefficient is +0.8, then each additional reception in a previous year means we expect 0.8 more receptions in the current year; this is in addition to a ‘base rate’, or intercept in the model, and just identifying the marginal impact of additional receptions. Generally, ‘sticky’ stats will have coefficients close to 1.0, since this just means that we predict the statistics to replicate themselves over time. ‘Non-sticky’ stats will be closer to zero, which just means that previous performance is less impactful when predicting future performance.

We don’t see coefficients above 1.0, thanks to the simple fact that, on average across all players and career points, performance will decline over time. This also makes sense when we think stochastically: a coefficient of, say, 1.5 would imply that we predict receptions to increase each year, compounding on and on without end!

Wide Receivers

We can begin with wideouts; the below chart tells us the coefficients for their major statistics. For example, the ‘rec’ bar tells us that each additional reception in a previous year means we should expect about 0.5 more receptions this year. Remember, values near 1.0 indicate ‘stickier’ statistics.

The FootClan
Unlock Exclusive Tools + Bonus Episode
Join the FootClan

It’s not surprising that yards and targets are on the ‘stickier’ side. Targets especially are often discussed on the podcast as ‘skill-based’, so it makes sense that they persist over time. What is surprising is that air yards, or the amount of yards that the ball travels in the air to a WR on a pass (complete or incomplete) is the stickiest statistic. This bodes well for D.J. Moore and Terry McLaurin, who landed, surprisingly, at 5th and 6th in air yards among receivers last season with about 1630 each (yes, we can take a moment to lament the untapped potential thanks to inconsistent QB play). I expect this to be pretty sticky for the 2022 season, which means that D.J. and Scary Terry should continue to see massive volume through the air.

Unsurprisingly, receiving touchdowns are the least sticky stat: there’s a lot of randomness behind who actually catches the touchdown on a drive, so these numbers don’t persist as well. Probably the most notable player that this affects is Adam Thielen, who defied Father Time with 10 scores through the air last year, tied for 6th at the position. His yardage, though, sat at 726, far below players with similar touchdown production. The trends say that Thielen should expect regression this season.

Now, don’t hear what I’m not saying. We can never blindly cling to what a model suggests. After all, these are average results for the average player, and represent plenty of uncertainty to boot. Adam Thielen may very well be an outlier: he’s great on short, red zone routes and has incredible chemistry with Kirk Cousins (not quite breakfast pals like Stafford-Kupp, but perhaps brunch?). There’s plenty of reasons to think that Thielen can defy the trends for another year, and it’s important to not abandon hope because of some model output. Still, it is important to keep the trends in mind, and to temper expectations for Thielen repeating in the TD category.

Running Backs

For running backs, we break the stickiness metrics down between receiving work and rushing work:


Again, perhaps surprisingly, air yards is by far the stickiest stat. This bodes well for Austin Ekeler, who recorded the 2nd highest total at the position. Cordarelle Patterson was, unsurprisingly, the air yards leader, but this was more a function of his usage as a wide receiver. On the flip side, though, receiving touchdowns for RBs are extremely not sticky, which is bad news for the aforementioned Austin Ekeler, who posted eight in 2021. The takeaway is something you would probably expect: Ekeler should get a ton of receiving work this upcoming year, but it’s unlikely that he repeats such a prodigious scoring rate after catching the ball.

The FootClan
Love the show? Join our community!
Join the FootClan

In terms of actually running the ball, usage, measured by attempts, is the stickiest stat. Antonio Gibson sneakily recorded the 4th highest carry total in the league last season, so we would predict a high workload in 2022 (even with the return of J.D. McKissic). Rushing touchdowns are similarly as ‘not sticky’ as receiving touchdowns for WRs, which tells a cautionary tale for Damien Harris (T-2nd in the league with 15 last year), especially after the New England Patriots continued to invest draft capital in the running back position.

Quarterbacks & Tight Ends

We can lump the “onesie” positions into one section:

What’s really interesting for Quarterbacks is that almost all of the statistics are particularly not-sticky. Touchdowns are the second stickiest stat, but they have a coefficient below 0.5, which is about the same as the coefficients for RBs and WRs (where it was the least sticky stat). This is an interesting result, and while it could be affected by the smaller sample size (less QBs than other positions), it’s a good reminder that QBs in general are difficult to project year over year. The coefficient on interceptions is basically zero (funnily enough, it’s negative), which means on average a previous year’s interceptions are of no use (or even have a negative relationship) in predicting this year’s interception rate. That’s good news for Matthew Stafford and Trevor Lawrence, who paced the league with 17 picks apiece.

One potential (unsurprising) implication is that it’s probably not worth taking a QB relatively early unless it’s one of the true alphas at the position. Outside of a couple of elite QBs, it’s a difficult position to predict in terms of actual end of season stat lines.

Tight ends have a similar profile to WRs, although their receiving touchdowns are even less sticky. That means we should be even more cautious of players like Hunter Henry and Dawson Knox, who had solid TE seasons because of touchdown amounts far beyond expected. On the flip side, Cole Kmet, who was heavily involved with the Chicago Bears‘ offense but caught zero touchdowns, should see some positive regression.


Hopefully this article gave you a quantifiable sense of what is ‘sticky’. The results are mostly unsurprising: yardage and usage are stickier over time, while touchdowns across the board are less sticky. However, there are some surprising results. Air yards are very sticky across positions, and receiving touchdowns for RBs are very not sticky. What’s more, nearly all of the statistics for QBs are not sticky over time! Keep this ‘persistence’ in mind when your drafts roll around, and always be on the lookout for regression, both positive and negative.

The FootClan
Unlock Exclusive Tools + Bonus Episode
Join the FootClan


Let me know what you think on Twitter.

Leave a Reply

Your email address will not be published. Required fields are marked *