Buy Leone WR and TE Underperformers

The Buy Leone WR and TE models are machine learning algorithms that use a player’s recent opportunities to predict future performances (PPR points per game).

More specifically, the models look at every target a player has seen over the last six weeks. The model provides an expected fantasy-point outcome for each individual target based on a variety of factors, most importantly:

-Team Scoring Expectation
-Air Yards
-Field Position

Those expected fantasy points become the primary input into a model that weights recent performance at a higher weight than past performance and incorporates aggregate per-game totals, such as Air Yards, targets, and fantasy points.

Our goal is to have an objective process to find the players whose opportunity has out-paced their actual results. Through that, we’re identifying players who have a high likelihood of beating their last three-game average moving forward.

In order to isolate the most potentially profitable players, we are comparing the Buy Leone model prediction against a player’s rolling three-game fantasy-point average (going back no further than the previous six NFL weeks) and filtering out all players who don’t meet the following criterion:

-Buy Leone prediction greater than rolling three-game fantasy-point average
-Minimum eight fantasy-point prediction via Buy Leone model (seven for tight ends)
-Minimum eight fantasy-point projection via ETR’s weekly projections (at the time of publication, seven for tight ends)

This model will improve as the year goes on. Currently, the model is using only the last four NFL weeks, excluding Week 18 of last season, and weighted very heavily to Week 1 of the 2023 season. As we get further into the season, the model will use the last six NFL weeks and will weigh most recent weeks a little less heavily (but still at a higher rate than past weeks). The WR and TE results are combined below, but these are two separate, albeit very similar, models.

It’s important to keep in mind that the Buy Leone model by design minimally accounts for player talent and surrounding team talent. This is intentional to keep the model as objective as possible. Still, it’s important to keep in mind that:

-Past player opportunity doesn’t guarantee future player opportunity (this is accounted for in the machine learning process)
-Player, team talent, and upcoming matchup are all important factors that may be undervalued by the model

As a result, Adam Levitan is leaving his context-adjusted takeaways and commentary below the model results.

Essentially, this is a list of guys who haven’t performed as well as their past opportunity indicates. Fantasy players who aren’t paying attention might be too down on their future prospects, and we think it’s good context to keep it in mind for DFS, trading, and more.

Enough housekeeping — here are the model results:

KEY

Buy Leone Proj – Mike Leone’s machine learning algorithm’s projected future PPR PPG, based primarily on recent player opportunity
Last 3 FPPG – A player’s average PPR PPG over their last three games
Under Performance – The difference between a player’s Buy Leone Proj and their Last 3 FPPG
*DFS Proj – A player’s weekly PPR projection based on ETR’s DFS projections