Week 7 Recap
Week 7 was a very good one for the model. While the players at the top of the model underperformed for a second week, those further down the list put up respectable and, in the case of Marvin Jones, spectacular weeks. Jones at just 1.3 percent ownership is precisely the type of player the model is trying to uncover.
Overall the hit rate (defined loosely as performance exceeding projected based on opportunity) for Week 7 was just over 40 percent, but could have easily been higher if players like Mike Williams weren’t overthrown by 4 feet in the end zone, and if Ted Ginn wasn’t tackled near the goal line after an underthrown ball by Teddy Bridgewater. Just saying, it’s not that I’m tilted at all.
Here are the full results for Week 7:
|Sterling Shepard||NYG||13.1||OUT – Concussion|
|Parris Campbell||IND||10||OUT – Abdomen|
|Christian Kirk||ARI||8.6||OUT – Ankle|
The Buy-Low model uses target share and air yards to estimate a player’s expected production in the passing game, then highlights the players that underperformed relative to expectation. The key insight behind the model is that opportunity is sticky and production (in the form of catches and touchdowns) is not. You want to buy the signal and fade the noise, and the model helps us do just that.
The out of sample r-squared for the model for this week is 0.57 (down 1 points from 0.58 last week).
Editor’s Note: Before using the model, we strongly suggest everyone read Josh’s article introducing the concept here. We also recommend you listen to his interview with Adam Levitan in Episode 4 of the ETR podcast.
In general, pay most attention to the projection column as it reflects the value of the opportunity each player received. The next piece of information you should weigh is the size of the difference between what the model says a normal game from this player should be given his opportunity, and his actual performance in the recent past. The larger this difference, the greater the chance that the public will be fading the player, making him low-owned. And while we might be tempted to infer that larger differences might lead to a stronger “rubber band” regression effect, it’s typically the case that what dominates is the opportunity.
* Projection = The full-PPR projection the model gives for a player for the rest of the season based upon his opportunity in the previous three games.
* Actual = A player’s average PPR points per game over the past three games.
* Difference = The difference between projection and previous week result in full-PPR fantasy points.
Teams on BYE this Week are: the Baltimore Ravens and the Dallas Cowboys.
|Player||Team||Projection||Actual||Difference||Main slate DK salary|
- Mike Williams returns at the top spot on the buy low for a third week in a row. Williams nearly had a touchdown Sunday as the Chargers continue to find new and exciting ways to lose, including overthrowing their 6’4” 220 pound wideout. The math says stay the course and keep starting him. At just $4000 on DK, fitting him in your tournament lineups shouldn’t be too difficult.
- Mike Evans is back on the list. Long time readers will recall that way back in Week 3 after appearing on the buy low Evans produced an explosive performance. In the past four weeks he has nearly equalled Chris Godwin’s share of team air yards and is trailing him by 2 targets a game on average. Yet his PPR point total is 37.5 to Godwin’s 97.8 over those four games. Positive mean reversion appears to be on the horizon.
- Jamison Crowder was not the high-floor, high-upside play I envisioned in Week 7 against New England, but he might be Week 8. Sam Darnold looked lost and broken, but things can change quickly in the NFL. Recency bias will be on your side, and at $4800 the price is right to take a shot on either Crowder or teammates Robby Anderson and Demaryius Thomas. Anderson also made the list, costs just $100 more, and actually has a higher projected PPR total based on his recent opportunity.