The Air Yards Model countdown continues. Up next: The Jacksonville Jaguars Allen Robinson.
Well-validated models are the best and most unbiased forecast of NFL receiver production there is.1 However there are things that they cannot capture that are important to consider.
Changes to team composition, injuries, coaching changes, and more can affect the accuracy of a pre-season model that is based purely on numbers. Finally, many models, including mine, do not give more than an expected value forecast for a player. Point estimates are great, but it’s also crucial to understand the range of outcomes.
In this series, I’m counting down the model’s top-36 wide receivers and publishing forecasts, along with a discussion of the other important factors the model doesn’t know about. Full projections will be released next month.
Air Yards Model Projection
Projected PPR Points per Game: 15.6
Projected WR Rank: 25
Actual ADP:2 16
Model Difference: -9
Team Personnel Changes
In the 2017 NFL draft, the Jaguars added three new offensive players. In the first round, the Jaguars selected running back Leonard Fournette to bolster their running attack. They followed with troubled but talented wide receiver Dede Westbrook in the fifth and wrapped things up with full back Marquez Williams in the seventh. The high pick of Fournette highlights the Jaguar’s move to a more run-oriented offensive philosophy and he looks to be a lock for early down work in 2017.
In free agency, the Jaguars focused almost entirely on the defensive side of the ball. Their two offensive signings were C Demetrius Rhaney from the Rams and TE Mychal Rivera to add depth behind projected starter Marcedes Lewis. None of the new Jaguars look to impact Robinson directly, but the increased emphasis on defense and ball control will almost surely have an impact on the number of targets Robinson will see in close games.
All discussions of Jacksonville must start with game script. The Jaguars have been consistently bad since 2007, posting a negative seasonal game script each year over that span.
Being behind often means abandoning the run, and that is precisely what the Jaguars have done each season. Their pass ratio over the past five years has been above league average by about 5 percent, a large number.
Jacksonville’s plays per game are league average, but the extreme pass split means that they throw more than the average NFL team. Last season they attempted 626 passes.3 The ARIMA model projects a drop in expected passes to 559 on the season and all signs seem to support this forecast. Early in games and whenever the score is close, it is clear the Jacksonville coaches and front office would prefer that the offense runs instead of passes.
The million dollar question is: how often will the score stay close?
Bovada Sports Book has the win total for the Jaguars at just six games indicating a negative seasonal game script. For most teams this would mean we could confidently project a high pass ratio, but we probably need to temper this with the Jaguars. Their defense is much improved and stocked with both young and veteran talent. The defense could keep the Jaguars close late into more than a few games. If this happens it will reduce the chances for the monster 45 pass comeback attempts that have fueled both QB Blake Bortles and Robinson’s fantasy scoring in the past.
I’ll accept the model’s recommendations. Here are the various passing numbers it projects.
Floor: 514 attempts
Ceiling: 604 attempts
Expected: 559 attempts
Since coming into the league, Robinson has never seen less than a 24 percent target share. This makes his forecast easy. There is no reason to think Jacksonville is looking to reduce Robinson’s target share, despite their desire to run more effectively and insulate the team from the inconsistent play of Blake Bortles. Let’s project his targets using the team totals we came up with above.
Floor: 123 targets
Ceiling: 145 targets
Expected: 134 targets
If you combine catch rate and yards after catch (YAC) into one metric you get something I call Receiving Air Conversion Yards or RACR. Over his career, Robinson has been average at converting a yard thrown at him into a receiving yard. He’s not particularly great anywhere on the field, but he does have a talent for coming up with yards on exceptionally deep passes.
Robinson’s YAC by depth is troubling. Like many big wide receivers, he doesn’t consistently create after the catch, which means drops and off target passes (which are mostly random) will have a tremendous impact on his fantasy output in a given week.
Nowhere is this more obvious than when we look at Robinson’s catch rate on deep targets from 2016. Past 25 yards deep, he failed to catch a single target last year.
Pro Football Focus has Robinson with 31 targets of 20 yards or more. This means 22 percent of his targets were deep looks. He was credited with only one drop, while just five of the throws in his direction over 20 yards were deemed by the PFF charters to be catchable. On 84 percent of his deep targets, Robinson had no chance.
Robinson’s career aDOT is 13.3, and his career RACR is 0.56, which is average. It means he converts each yard thrown at him into just over half of a receiving yard. Let’s calculate his receiving yards.
Floor: 123 targets * 13.3 aDOT => 1636 air yards * 0.56 RACR => 916 receiving yards
Ceiling: 95 targets * 13.3 aDOT => 1929 air yards * 0.56 RACR => 1080 receiving yards
Expected: 90 targets * 13.3 aDOT => 1782 air yards * 0.56 RACR => 998 receiving yards.
Robinson’s career catch rate is a woeful 53 percent. Some of this is driven by the deep targets he gets, some of it is driven by Bortles, but some of it is also that Robinson has dropped more short passes than you might expect. Let’s calculate his receptions.
Floor: 123 targets * 0.53 Catch Rate => 65 receptions
Ceiling: 145 targets * 0.53 Catch Rate => 79 receptions
Expected: 134 targets * 0.53 Catch Rate => 71 receptions
Finally, let’s calculate TDs. Robinson’s most elite quality is his ability to use his big body and athleticism to elevate in the end zone and come down with contested touchdown grabs. His catch rate on end zone targets is excellent, and he should continue to be targeted there.
Robinson’s career TD rate is 5.3 percent.
Floor: 107 targets * 0.053 TD Rate => 7.1 TDs
Ceiling: 143 targets * 0.053 TD Rate => 8.4 TDs
Expected: 137 targets * 0.053 TD Rate => 7.7 TDs
Robinson’s final back-of-the-napkin projection is:
Let’s compare it to the Sim App:
The Sim Score App, which uses historical comps to derive its projections, again finds a larger spread than our back of the envelope calculations. The Sim Score app does validate our general range, however, while offering a glimpse of both the potential upside and downside in Robinson’s profile. The air yards model’s expected PPG total comes in well above our back-of-the-napkin calculation, which is interesting but can probably be discounted somewhat given the things the model doesn’t know about the changes on Jacksonville, both in personnel and offensive philosophy.
Allen Robinson is an air yards stud. In his first few seasons in the NFL he has gotten a ton of yards thrown his way. Unfortunately many of those yards were wasted because the throw was off-target or dropped. This was especially true in 2016. Because of this, Robinson has a limited number of chances to pull down a catch and create after it, depressing his RACR. While Robinson’s target share is likely to remain stable, the Jaguars will probably throw less in 2017 than they did the year prior. I don’t think they will run as much as some analysts are projecting – the Jaguars are still not a good team – but Robinson’s raw target numbers will likely suffer and the market hasn’t fully priced this downside in. Robinson is being over drafted at his current ADP and is unlikely to return value.
- The air yards model combined four separate machine learning models into one ensemble model. The models used included a Bayesian neural net, a nearest neighbors algorithm, a random forest algorithm and a tuned generalized linear model. Preprocessing steps included data normalization (centered and scaled), the elimination of near zero variance predictors, and principal component analysis transformation. The model data used were 838 observations of 35 predictors, including age, weight, previous season player volume and efficiency data, 3 year weighted player efficiency data, and previous season team passing volume and efficiency data. The dependent variable was fantasy points per game. The data were split 80/20 into training and test sets. 10 fold cross validation was performed. The measured R^2 on the held out data was 0.66. (back)
- as of end of July 2017 (back)
- League average is around 575. (back)