Players Due for Regression (Part 1)
Micah Tweeten
|
Defining Regression
Fantasy football is a game of statistics and only concerns itself with statistics. Take the Minneapolis Miracle for example: fantasy football doesn’t care that Stefon Diggs leaped in the air, safety Marcus Williams missed low on the tackle and went underneath Diggs, who then used his hand to keep himself up and in bounds before racing 61 yards for a touchdown as time expired for one of the most dramatic plays in NFL history. It only cares that Stefon Diggs had 61 receiving yards and a receiving touchdown. That’s it. It’s a game that cares about how many times you scored and doesn’t care at all about how you scored. And because fantasy football is such a statistic-based game, we can use statistical analysis in order to look at and find players that will likely not repeat the efficiencies (or inefficiencies) of the season before. One of the metrics we can use is called regression. Regression is simply defined as returning to the mean or average – this may mean a player was more efficient than what they should have been and will likely perform worse than the previous year, but it can also mean positive regression which means that the player performed worse than what we would have expected and therefore will likely perform better than the previous year. (Note: Positive regression is not the same as progression. Progression is defined as developing to a more advanced state whereas positive regression is just returning to a previous state in a beneficial sense). This helps us be able to find inefficiencies in the market while we are drafting and we can snag a player that some people are avoiding because they didn’t perform well the previous year, or we can avoid players that are very unlikely to repeat their 2017 season.
Defining Opportunity
I was curious as to what players from the 2017 season produced more efficiently than they should have based on opportunity and look to negatively regress in the 2018 season. Before I share my findings, let me define opportunity. Opportunity is a way to look at different things like targets, rushing attempts, pass attempts, and where those are happening on the field so that you can evaluate whether or not that player has the opportunity to not only gain yards but score touchdowns, which is where we get the most points in fantasy football. A good example of evaluating opportunity is in this example: Say we have take the Vikings backfield from most of last year (Latavius Murray and Jerick McKinnon). We give Murray ten carries inside the opponents five yard line and then we give Jerick McKinnon 25 carries from the midfield line. That would seem to indicate that McKinnon is to get a significantly higher amount of touches in the Vikings offense (2.5 times the amount of work as Murray). But if we look at opportunity, we would expect Murray to score more touchdowns than McKinnon. Within those ten carries, Murray might score three touchdowns and rush for 35 yards (or 3.5 yards per carry). This means we would expect him to score somewhere in the range of 21.5 points for those ten carries. Now, let’s say McKinnon rushes for an extremely exceptional six yards per carry for 150 yards (whose yardage is now nearly five times that of Murray). But McKinnon would be rather unlikely to score from midfield in those 25 touches (he ran the ball for 10+ yards only 8% of his carries last year, meaning only 2 of his 25 carries would be 10+ yards and unlikely that he goes for 50 yards). Given he doesn’t score in any of those tries, McKinnon would get 15 points for those 25 carries. Despite getting 15 fewer carries than McKinnon, Murray scores 143% of the amount of points that McKinnon scores.
Application
It’s this understanding of opportunity, volume, and expectation that helps us figure out a player’s regression. We can expect a certain amount of yards and touchdowns from a player based on the amount of touches he has and where he’s getting those touches. It was using this model that I found sixteen different players (eight negatively regressing and eight positively regressing) that I think we should not expect them to perform like last year for the upcoming 2018 NFL season. Please note, players due for negative regression are not players that I think are busts or that aren’t worth drafting or that aren’t still elite tier players at their position. Rather, they are players that we are currently either drafting at their potential ceiling based on last year’s results as opposed to taking them where we think they will realistically land this season. So with that, here is part one (of four) with some of the players that we can expect to negatively regress in the upcoming season.
Deshaun Watson, QB, Houston Texans
2017 Stats: 7 games, 126/204, 1699 yards, 19TDs, 8 INTs
2017 Finish: 164 points, 23.4 PPG, QB26 (QB1 when healthy)
Deshaun Watson was the player taking fantasy football by storm last year – if someone in your league hadn’t drafted him, people were rushing to grab him from the waiver wire after his 26 point performance in week three. And if you happened to get him, things only got better from there. According to FantasyData.com, Watson put up three performances of 32+ along with another performance of 23 before going down with a torn ACL. This four game outburst, which saw him throw for no fewer than three touchdowns in a game, helped launch Watson into the elite tier of quarterbacks for this upcoming 2018 fantasy football season. But is Watson ready to repeat that performance, or is he going to come hurtling back to earth and ready to disappoint any player who drafts him at his current fourth round ADP? Based on the metrics, Watson looks poised to heavily regress statistically this upcoming season – but that doesn’t mean you still can’t take him early and succeed.
Touchdown Machine
Just a year ago, Watson threw for 19 touchdowns in just seven games (and only started six of those games). But my expected touchdown metric, which looks at 10 years of data in the NFL and analyzes both the average expected touchdown rate for an NFL quarterback in a given situation as well as the players career (as long as the sample size is big enough), said that Watson should have only thrown for just over nine touchdowns last year – a ten touchdown decrease from what he did. If we extrapolated his statistics last year over a 16 game season, we would have expected him to throw 43 touchdowns through the air last year – nine higher than the league leader Russell Wilson. Instead, what this metric says is that he should have actually thrown just under 22 touchdowns (or tied for 12th best in the league last year with Cam Newton, Dak Prescott, Case Keenum, and Derek Carr). While still an above average touchdown mark, it’s not nearly the high level number that we would expect given his seven games last year.
Interception Prone
Another number of Watson’s could likely regress this year as well – but regressive positively. Watson threw eight interceptions in those seven games last year, but again, based on expected interception metrics (which is like my expected touchdown metric, but for interceptions) shows that he should have only thrown 4.95 interceptions in those seven games. Now we can either read this to be that he doesn’t make the best decisions (and there is definitely some of that in play when you watch his tape) or that he doesn’t have the most accurate or strongest arm (also a very real possibility), but it also could play into the fact that he was partly unlucky with his interceptions and can also grow this offseason in decision making to avoid making some of those mistakes.
Elite or Unrepeatable?
Now, despite the interception numbers possibly going down, the strong possibility of efficiency regressing scares me because not only was his touchdown efficiency otherworldly last year, but his yards per attempt was also exceptionally high. Watson averaged 8.3 passing yards per attempt last year, second behind only Jimmy Garoppolo last year (another quarterback who will likely regress significantly in the yardage portion of things). That resulted in the 21st best season in the last 15 years in terms of yards per attempt for quarterbacks who had at least 200 passing attempts in a given season. The other quarterbacks on the list that passed Watson were players that either won the MVP that year (Matt Ryan in 2016, Peyton Manning in 2013, and Aaron Rodgers in 2011) or were players like the aforementioned Ryan, Manning, and Rodgers (who repeated several times) as well as Tom Brady, Drew Brees, Philip Rivers, and Eli Manning. Again, you can read this as meaning Watson is a future franchise quarterback, but likely what it means is that his numbers were so high based on efficiency through the air, that it will likely be unrepeatable this year.
So what does all this regression mean for Watson? Well, if we take his season last year, account for regression in his stats (both positive and negative regression), and extrapolate it to a full 16 game season, Watson would have finished as the QB5 – which really doesn’t sound that bad despite the fact that he was very clearly the QB1 when he was playing last year. That still presents as a top-tier quarterback option in fantasy – but by drafting him at QB2 we are drafting him based on his best case scenario where he continues his tear from last year with the same efficiency. Instead, we need to plan for regression on Watson and we will likely see him finish between QB4 and QB7 this year. He still has a solid floor due to his ability on the ground, but he’s a player that will regress significantly and if we just draft based on his performance last year, we will be due for some disappointment throughout the year.
Carson Wentz, QB, Philadelphia Eagles
2017 Stats: 13 games, 265/440, 3296 yards, 33 TDs, 7 INTs
2017 Finish: 270 points, 20.8 PPG, QB5 (QB3 when healthy)
Carson Wentz was a quarterback that I loved last year. Despite being taken as the QB18 in drafts last year, I was trying to grab him in every league I could in the 10th, 11th, or 12th rounds as my starting quarterback. Regression and progression statistics showed that if Wentz made the jump that a normal second year quarterback does in fantasy, he would finish as a low tier starting quarterback for fantasy purposes (in the QB9-QB12 range). As a result, he was one of the players that I had singled out as a potential breakout candidate and exceptional late round value for 2017. But despite being ahead of the curve compared to most on Wentz’ breakout, I didn’t see his 2017 season coming. Wentz was the NFL MVP frontrunner and QB3 before he went down last season with a torn ACL, and still finished as the QB5 in standard leagues.
League Leader
Wentz performed well with a solid amount of yards and a high touchdown total that was just one away from the NFL leader (Russell Wilson), all while keeping his interception total very low throughout the year in an all-around good performance for the eventual Super Bowl champion Eagles. However, Wentz performed at a rate much higher than what was to be expected of him based on the amount of throws he was having throughout the year. Part of this was due to the high quality of offense of the Eagles, but another part of this was due to an over-efficient performance by Wentz throughout the year. Wentz ended up throwing 33 touchdowns last year in just 13 games or almost three touchdowns per game. Wentz realistically should have only thrown for just over 20 touchdowns throughout the year based on the previously stated expected touchdown metric, which is a substantial difference of nearly 13 touchdown throughout the year. That difference between expected touchdowns and actual touchdowns was the highest out of any quarterback last year and would have resulted in Wentz scoring four less points per game if he had been on par with the metric.
Interception Regression
Now, not only did Wentz outperform what he should have in the touchdown category, but he also outperformed what he should have in the interception category. Wentz threw for only seven interceptions just a year ago but really probably should have thrown closer to 11. Now, I understand that four interceptions is not a large number and four interceptions is within the range of outcomes (although on the very low end), but we should expect it to likely go up over the coming year.
Key Offensive Changes
In addition to the statistical regression, Wentz also lost tight end target Trey Burton this past offseason as well as offensive coordinator Frank Reich and quarterback coach John DeFilippo, who is now the offensive coordinator of the Vikings and was a major part in Wentz’ progression between years one and two. To replace Burton, the Eagles drafted tight end talent Dallas Goedert from South Dakota State to replace him in the second round. I was a big fan of Goedert going into the NFL draft and thought he could be a quality replacement for Kyle Rudolph in the coming years if the Vikings could have landed him in the first or early second, so I think the replacement of Burton was pretty well handled through Goedert. But we’ve also seen statistically in the NFL that rookie tight ends don’t perform all that well. We can’t expect Goedert to be the huge impact player on offense that Burton looked like he was becoming towards the end of the year last year. In free agency, the Eagles also resigned Darren Sproles, the seemingly never-aging running back who can catch passes out of the backfield. Sproles, along with Corey Clement, will be great complementary pieces to running back Jay Ajayi throughout the year, but as to whether or not they’ll significantly help Wentz’ statistical passing output is unlikely.
Room to Grow?
One thing we could see Wentz improve on this year is rushing touchdowns. Wentz showed great mobility in the pocket and when scrambling last year and we would have expected him to score nearly three times on the ground based on his carry total and where he got carries from last year, but he didn’t score once in the entire season. I would expect him to punch a few in on the ground throughout this upcoming season, which helps keep his floor a bit higher.
If we were to take the aforementioned regression (both positive and negative) of Wentz and extrapolate it throughout a 16 game season, Wentz would have finished as the QB12 last year instead of QB5. That still means Wentz is a low-end starting quarterback in leagues with twelve or more teams and is still a very high-end backup player for eight or ten team leagues. Wentz is still a very usable player in fantasy and should be drafted, but unless he and the Eagles offense continue their efficiency levels of last year, he will not be the high-end quarterback we expect him to be. Right now Wentz is being taken as the QB5, which is where I think his ceiling may be for the upcoming season, so if you are taking Wentz there, you need him to perform to the same level as last year, or you may be disappointed with the results.
Jared Goff, QB, Los Angeles Rams
2017 Stats: 16 games, 297/478, 3809 yards, 28 TDs, 7 INTs
2017 Finish: 252 points, 15.8 PPG, QB12
I’ve never been a big believer on Jared Goff in the NFL. Going into the 2016 NFL draft, I thought he was a second-round talent and not even the top quarterback in the class (that spot belonged to Carson Wentz) and I was disappointed that the Rams took him first overall. As he was performing his rookie year, he seemed to perform to what I thought he would be in the NFL more than what we saw last year. I was surprised by his ability passing the ball last year, not from a talent standpoint because he’s shown the arm strength and ability on that end, but on the mental and mistake making portion of his game. I believe a huge part of that is due to Sean McVay running the Rams offense, and helping Goff through the play-reading aspect before the snap. (And we saw some of that fall apart as defenses made adjustments after the headsets shut off before the play clock expired.) But despite my opinion of Goff, he performed well last year and ended up finishing as the QB12 in standard leagues, good enough as a starter in twelve team leagues.
A Surprising Performance
However, Goff looks to take a step back this season as his efficiency was beyond what it should have been. Goff threw for 28 touchdowns last year through 16 games or almost two touchdowns per game. I realize that this doesn’t sound like a highly efficient number, and it’s really not other-worldly, but it was still good enough to place Goff in the top ten for passing touchdowns last year in the NFL. But despite that top ten finish, Goff should have thrown for eight less touchdowns than he did as Goff was expected to throw just over 21 touchdowns last year based on the opportunities he had. I think part of his success was attributed not only to his play through the air, but also Gurley’s play on the ground and while I think the “run setting up the pass” mentality of NFL offenses is highly overrated, I do think that Gurley’s ability both on the ground and through the air helped to take some of the pressure off of Goff to make quality throws.
Avoiding the Plaguing Problem
In addition to the excess of touchdowns, Goff also seemed to throw less interceptions then what would be expected. Goff avoided making a ton of mistakes and only threw seven interceptions last year, but the metric would expect that he would throw over 11 interceptions throughout the course of the season. Again, like Wentz, this isn’t outside of the range that we can expect him in, but it is definitely on the lower end of efficiency and seem to be an outlier in Goff’s career in both college and the NFL. Prior to last year, Goff had thrown 37 interceptions in college and the first year of his NFL career, and if we extrapolate those numbers to a 16-game season, that’s an average of 13.77 interceptions in any given season – nearly double the seven interceptions he threw last year. I would expect to see Goff back in that 11 to 15 range again this year.
If we took Goff’s regression and extrapolated that data across a 16-game season, we would have expected Goff to finish as the QB23 last year, not the QB12, which is where he finished at. I think this shows that Goff’s floor is much lower than some of us may expect after his quality year last year. I don’t think that Goff is necessarily unrosterable as a quarterback, but I do think we need to exercise some caution in understanding where Goff could land this year in the chain of quarterbacks. I’ve seen some (rather ridiculous) analysis on Goff and that for where he’s going in the tenth round, he’s a much better value than players like Kirk Cousins and Matt Ryan because they project he’ll score the same as Cousins and Ryan. In these cases, they are expecting the same unlikely efficiency from Goff and looking near his ceiling while looking closer to Ryan and Cousin’s floors. I think there are better options that are coming at or after Goff’s current ADP of 10.02 and as the QB13 (see: Mariota, Marcus or Smith, Alex).
Todd Gurley, RB, Los Angeles Rams
2017 Stats: 15 games, 279 carries, 1305 yards, 13 TDs, 64 receptions, 788 yards, 6 TDs
2017 Finish: 317 points, 21.1 PPG, RB1
Todd Gurley is an interesting candidate for regression because, as will be discussed more in the next paragraphs, he was unsustainably efficient last year – but even if he was ‘average’ in terms of efficiency and his fantasy total dropped to where we would have expected with average efficiency, he still would have finished as the RB1 last year. That’s how good of a year he had and how much opportunity he was given in the Rams offense. But can Gurley come close to repeating the opportunity and production that we saw from him last year? Perhaps, but Gurley is due for some substantial regression in nearly all aspects of his performance.
Regressing the Regression?
In some ways, Gurley positively regressed last year back to his rookie year performance after a slumping sophomore season. But not only did Gurley regress, but he overshot that regression and is now due for negative regression again this year. Gurley’s yards per carry returned back to a healthy 4.7 after dipping to 3.2 in 2016. This landed him about where we would have expected after an even better 4.8 yards per carry in his rookie year. But while his yards per carry came back to the average, Gurley’s touchdown rate soared. In 2016, Gurley rushed 278 times and scored six times on the season, a bit lower than we would have expected (expected touchdown number was just shy of 8). Last year, Gurley rushed just one more time on the season as a whole and scored over twice as many times – a league leading 13 touchdowns on the ground. Gurley was expected to score 8.7 times with the touches he received, meaning he scored nearly 4.5 rushing touchdowns more than we would have expected him to with the touches he received. Now, this could be within the upper echelon of his expected range as it was the second time in three years that he rushed for 10+ touchdowns in a year, but it seems to point to some regression in scoring on the ground – especially since a lot of other parts of his game seem due for regression.
A Historical Performance
The biggest piece of Gurley’s game that seems ready to regress is his work in the passing game. Never before have we seen a performance like Gurley’s by a running back in the passing game: Gurley’s 12.3 yards per reception last year was the 6th best performance by a running back in the passing game since 2003. Of the five running backs ahead of him, none were bellcow RBs like Gurley is and none caught more than 40 passes in that season while still putting up the Y/R that Gurley did. Gurley, on the other hand, had 64 receptions last year while also carrying the ball almost 300 times. To add to the unlikelihood of Gurley repeating, none of the running backs were able to come close to repeating their efficiency in the next year. The closest was Tevin Coleman last year with a 2.8 Y/R decrease from the year prior. The other piece of Gurley’s receiving game that helped buoy him as a fantasy star was his ability in catching TD passes. In the two years prior, Gurley had never caught a TD pass, and while he was due for some regression (we would have expected 2.6 in his career before last year), he surpassed that number by catching six touchdowns last year when we would have expected him to catch 2.5. We can expect that number to drop back to how he’s performed in the past this season and he should finish with two or three receiving touchdowns in 2018.
If we take all of that regression and apply it to last year, like we touched on in the beginning, Gurley still would have finished as the RB1. But the difference is the amount of points he would have ended up with. Instead of the 317.3 points we saw him finish with, Gurley would have ended up with 265 points, just barely ahead of RB2 Le’Veon Bell. We can certainly justify Gurley’s current draft position (1.02, RB2), but we shouldn’t be expecting the same ceiling as he had just a year ago, especially if Goff and the offense as a whole also regresses from where they were last year. All this is to say, if you take Gurley in the first round, draft him because of his solid floor due to opportunity and upside as a dual threat running back, not because of his perceived ceiling from last year’s performance.
Conclusion
So with that, those are four players that I think we need to watch for negative regression this year in fantasy football. Make sure you watch for part two to find out more players due for negative regression, coming later this week!
Think I’m wrong about a player, have other players you think are due for regression this year, or just want to ask a question about this upcoming fantasy football season? Leave a comment below or email me at [email protected].