The Air Yards Model countdown continues.
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: 15.2
Projected WR Rank: 26
Actual ADP:2 48
Model Difference: + 22
Team Personnel Changes
In the 2017 NFL draft, Minnesota added four new offensive players. Starting in the second round, the Vikings selected running back Dalvin Cook to help replace the departed Adrian Peterson. They followed with wide receiver Rodney Adams in the fifth and wrapped things up with tight end Bucky Hodges and WR Stacy Coley in the sixth and seventh rounds. Of the four, Cook has the best shot to make a fantasy impact in 2017.
In free agency the Vikings made a number of moves to bolster their offense. The Vikings added LT Riley Reiff and RT Mike Remmers, two middling linemen, great at neither pass blocking nor opening holes in the run game according to the graders at Pro Football Focus. At the skill positions they added RB Latavius Murray, WR Michael Floyd, and backup QB Case Keenum.
None of the signings look to negatively impact Thielen’s outlook, as Floyd and his Kombucha baggage is more of a depth addition than a threat to steal playing time from last year’s starters.
Last season the Vikings threw at a league average rate, the first time in seven years that they’ve done so. This was coupled with a neutral seasonal game script and near league average plays per game. Along with the loss of Peterson, this may indicate the Sam Bradford-led offense will continue to pass more than we might otherwise expect given their history.
In 2016 the Vikings attempted just 588 passes. The ARIMA model forecasts a large dip in 2017, down to an expected 527 pass attempts. For the reasons above, I think this is probably too low. Peterson is gone, and they have a a solid set of wideouts in Stefon Diggs and Thielen. They will pass more than they have in the recent past, though perhaps not as much as last season. I’ll peg it at 560 expected pass attempts. Here are the full set of ranges offered by the ARIMA model:
Floor: 482 attempts
Ceiling: 590 attempts
Expected: 560 attempts
The average NFL team attempts roughly 575 passes a season, so we’re projecting close to an average passing attack. If the Vikings underperform and find themselves with a negative seasonal game script, this could easily go up. Bovada Sports Book has the win total for the Vikings at just 8.5 however, so our best estimates should include a neutral game script.
Last season Thielen earned a target share of 16 percent. Perhaps contrary to expectations for a receiver with his level of grit, he only lined up in the slot 29 percent of the time, so he is not formation dependent. Further, his target share didn’t really fluctuate when Diggs was on the field and healthy vs. when Diggs was injured. His usage in the Vikings offense is largely independent of Diggs.
Projecting an increase or decrease in Thielen’s target share is perhaps the most difficult part of his forecast. He does not have a long history of being one of the focal points of a passing offense. He may simply be Plan B for the Vikings after LaQuon Treadwell failed to impress his rookie season. We’ll forecast a repeat of 2016 regardless of the underlying uncertainty. Here are his target calculations, based on the team totals we came up with above.
Floor: 77 targets
Ceiling: 95 targets
Expected: 90 targets
If you combine catch rate and YAC into one metric you get something I call Receiving Air Conversion Yards or RACR. Over his career, Thielen has been good at converting a yard thrown at him into a receiving yard, especially in the deeper parts of the field.
His catch rate mirrors his RACR (which is to be expected), but is far noisier. This suggests our forecast will be quite a bit more speculative for Thielen compared to other players.
Thielen’s career aDOT is 10.8, and his career RACR is 0.93, which is excellent. It means he converts almost every yard thrown at him into a receiving yard. Let’s calculate his receiving yards.
Floor: 77 targets * 10.8 aDOT => 832 air yards * 0.93 RACR => 773 receiving yards
Ceiling: 95 targets * 10.8 aDOT => 1026 air yards * 0.93 RACR => 954 receiving yards
Expected: 90 targets * 10.8 aDOT => 972 air yards * 0.93 RACR => 903 receiving yards.
Thielen’s career catch rate is an incredible 72 percent. There is a lot of error around this number though. Let’s calculate receptions.
Floor: 77 targets * 0.72 Catch Rate => 55 receptions
Ceiling: 95 targets * 0.72 Catch Rate => 68 receptions
Expected: 90 targets * 0.72 Catch Rate => 65 receptions
Finally, let’s calculate TDs. Thielen’s career TD rate is 4 percent.
Floor: 107 targets * 0.04 TD Rate => 3.7 TDs
Ceiling: 143 targets * 0.04 TD Rate => 4.6 TDs
Expected: 137 targets * 0.04 TD Rate => 4.3 TDs
Thielen’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, is in general agreement with our back-of-the-envelope forecast but has a much higher spread between Thielen’s upside and downside totals. The air yards model’s expected PPG total comes in at the very high end of the Sim Score’s range and completely obliterates our back-of-the-napkin calculation. This indicates that perhaps there is upside inherent in Thielen’s profile and situation that we aren’t completely factoring in.
Thielen is not a trendy pick this offseason. His teammate Stefon Diggs is a Rotoviz darling and gets most of the attention. There may be more upside in Thielen than his ADP gives him credit for, however. In his first full season of snaps he was hyper-efficient, and his role was rock solid. If it were to expand, or if Diggs were to be injured again, it is possible he could end the year with enough targets to thrust him into the high-end WR3 discussion. For a player with a positional ADP of 48, that is a lot of upside for very little investment.
- 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)