For those who are new to the feature, here's the deal: every week, I break down a topic related to regression to the mean. Some weeks, I'll explain what it is, how it works, why you hear so much about it, and how you can harness its power for yourself. In other weeks, I'll give practical examples of regression at work.

In weeks where I'm giving practical examples, I will select a metric to focus on. I'll rank all players in the league according to that metric and separate the top players into Group A and the bottom players into Group B. I will verify that the players in Group A have outscored the players in Group B to that point in the season. And then I will predict that, by the magic of regression, Group B will outscore Group A going forward.

Crucially, I don't get to pick my samples (other than choosing which metric to focus on). If I'm looking at receivers and Justin Jefferson is one of the top performers in my sample, then Justin Jefferson goes into Group A, and may the fantasy gods show mercy on my predictions.

And then, because predictions are meaningless without accountability, I track and report my results. Here's last year's season-ending recap, which covered the outcome of every prediction made in our seven-year history, giving our top-line record (41-13, a 76% hit rate) and lessons learned along the way.

Our Year to Date

Sometimes, I use this column to explain the concept of regression to the mean. In Week 2, I discussed what it is and what this column's primary goals would be. In Week 3, I explained how we could use regression to predict changes in future performance-- who would improve, who would decline-- without knowing anything about the players themselves. In Week 7, I explained why large samples are our biggest asset when attempting to benefit from regression. 

In Week 9, I gave a quick trick for evaluating whether unfamiliar statistics are likely stable or unstable. In Week 11, I explained the difference between regression and the gambler's fallacy, or the idea that players are "due" to perform a certain way. And in Week 12, I showed how understanding regression can allow us to predict the past as easily as the future.

Sometimes, I point out broad trends. In Week 5, I shared twelve years worth of data demonstrating that preseason ADP held as much predictive power as performance to date through the first four weeks of the season. In Week 15, I offered sobering data on why the best team usually loses in the fantasy football playoffs, and in Week 16, I explained why what works in the long run doesn't always work in the short run.

Other times, I use this column to make specific predictions. In Week 4, I explained that touchdowns tend to follow yards and predicted that the players with the highest yard-to-touchdown ratios would begin outscoring the players with the lowest. In Week 6, I explained that yards per carry was a step away from a random number generator and predicted the players with the lowest averages would outrush those with the highest going forward.

In Week 8, I broke down how teams with unusual home/road splits usually performed going forward and predicted the Cowboys would be better at home than on the road for the rest of the season. In Week 10, I explained why interceptions varied so much from sample to sample and predicted that the teams throwing the fewest interceptions would pass the teams throwing the most.

In Week 13, I explained that rookies were the only players whose production increased as the season went on and predicted that this year's rookie receivers would score more down the stretch. And in Week 14, I noted that large samples were almost always more predictive than small ones, and therefore "hot" players would likely regress toward their full-season averages.

The Scorecard

Our rookie receiver prediction closed with mixed results. On the one hand, the group as a whole saw a significant increase in its production and several young players have been veritable league-winners in recent weeks. Brian Thomas Jr. and Jalen McMillan have both scored more than 10 points per game higher than their prior average (Thomas has been the #4 fantasy receiver over the last four weeks). Rome Odunze, Xavier Worthy, and Ladd McConkey are also producing significantly more in recent weeks.

On the other hand, I predicted that this improvement would be broad-based, and it hasn't been. Ja'Lynn Polk, Adonai Mitchell, Malachi Corley, Jermaine Burton, and Luke McCaffrey haven't seen the uptick in playing time that is common among highly-drafted rookies late in the year; all five averaged 2.2 points per game or less.

As for our next prediction, some hot players have maintained their elevated pace and will likely be among the biggest league-winners in fantasy this year. Ja'Marr Chase, Keenan Allen, Josh Jacobs, Sam LaPorta, and Jason Sanders are all averaging more points over the last three weeks than they did over the four prior (their initial "hot" stretch). But more than twice as many of the "hot" players (12 out of 29) are underperforming their full-season average, so the group as a whole has regressed 85% of the way back to where they started.

Incoming Talent Regresses, Too

In 2018, I wrote about the perceptions that NFL careers were longer than ever before. Surprisingly, I discussed how they most certainly were not getting any longer (at least among the very oldest players), and how any perceptions to the contrary were mostly driven by a super-talented group of future Hall of Fame quarterbacks, headlined by Tom Brady.

In fact, there's no other position where careers are getting longer like they are at quarterback. In the last decade, eight different offensive linemen have started at least half a season at age 36 or older. Six different players did it in the year 2000 alone. There were seven different double-digit sack seasons by a 36-year-old player between 1997 and 2000. There has been one in the 17 years since. Even kickers aren't seeing any major improvements. From 2000-2009, the league averaged three kickers and punters per year over the age of 40. From 2010-2017, it averaged 2.5. (Old placekickers were slightly up, but old punters were way down.)

The evidence continued to mount against the "careers are getting longer" hypothesis to the point where last year, I wrote that the opposite was likely occurring: NFL careers were actually getting shorter. I think. Probably.

You see, studying trends in career lengths is devilishly tricky because of two simple facts:

1. Good players have longer careers than bad players.
2. The number of good players entering the league in any given year is not constant. 
   Already a subscriber? Login

   Continue reading this content with a PRO subscription.

   Join Now

Usually when we see a lot of "old" quarterbacks at once, it's not because of structural forces that are lengthening quarterback careers. It's because fifteen years prior we saw a huge influx in incoming quarterback talent. Twelve quarterbacks entered the league between 2001 and 2005 who would go on to pass for 20,000 career yards or more (Michael Vick, Drew Brees, Marc Bulger, Carson Palmer, Eli Manning, Ben Roethlisberger, Philip Rivers, Matt Schaub, Tony Romo, Alex Smith, Aaron Rodgers, and Ryan Fitzpatrick). Only four such quarterbacks entered between 2006 and 2010 (Jay Cutler, Matt Ryan, Joe Flacco, and Matthew Stafford).

There were a lot of implications of these trends. For instance, the 2018 season saw more quarterbacks from the 2001-2005 classes attempt a pass (11) than from the 2006-2010 classes (8) despite the latter group being five years younger on average. In 2019, the 2001-2005 class led with 8 quarterbacks to 4. In 2020, the older quarterbacks still led 6 to 5. (And Matt Schaub appeared in a game without attempting a pass, giving the older quarterbacks a seventh active player.) It wasn't until 2021-- sixteen years into the careers of even the youngest players in the first cohort-- that the 2006-2010 classes finally featured more active quarterbacks.

(And please note that Tom Brady doesn't even count for the 2001-2005 group-- he was drafted in 2000.)

Most people think of regression as some sort of balancing mechanism for luck at a player level. A receiver has scored "too many" touchdowns (or "too few"), and their luck reverts to form. Or a running back breaks a pair of long runs in a short span but is unlikely to do so again any time soon. This is true-- regression is the process of variance washing out of the system.

But it doesn't just operate at the player level or cover things we generally think of as "luck". Being a regression maximalist means seeing careers get longer or shorter and recognizing that these trends are also being driven by regression. It was unusual to see so many Hall of Fame quarterbacks enter the league in such a short time; there is no way it could be sustained, but the impact of that influx was still felt a decade and a half later.

Here's a chart of the average age of top fantasy performers since 2008. I've weighted ages by fantasy points scored-- so this year, 30-year-old Derrick Henry (the #2 RB in standard scoring) counts more in the average than 23-year-old Zach Charbonnet (RB30). (2.26 times more, to be precise.)

All of the cells are compared to all other years from the same position, with YOUNGER years shaded red and OLDER years shaded blue.

You can quickly see how chaotic and noisy these shifts really are. You can also see how it can be difficult to pick up on meaningful trends. Did running backs get significantly younger in 2017 and 2018 because the league suddenly favored rookies again? No, it was because the 2017 running back class was one of the best in history. (Similarly, the position has been getting older the last few years largely because... the 2017 running back class remains one of the best in history.)

Trends will still happen. The league will slowly shift towards players on their rookie contracts or toward veterans, especially as changes in the CBA alter the costs and benefits of each cohort. But these trends will always be obscured in the short term by random fluctuations in the underlying talent pool.

And just like all other random fluctuations, those year-to-year changes will tend to strongly regress toward the long-run averages over time.