Rest of Season RB Ranks – Week 7
Now that several players have six games worth of data for 2017, the model will be at its strongest point from here on out. That’s because I’ve shown before that six games of recent data is the most predictive in terms of predicting future performance, at least for a single-term model.
Adding other terms, we get a model that has an out-of-sample R-squared value of 0.45, meaning the model explains about 45 percent of the total variation in fantasy points per game from here to the end of the fantasy season (through Week 16). The terms in the Week 7 model are:
- PPR points per game Weeks 1-6
- Rushing yards per game Weeks 1-6
- Current week Depth Chart
- Receiving yards per game over the player’s last 16 games played
This captures several things. First, it captures form this year, especially on the ground. Second, it captures long-term receiving form. Most running backs don’t transition from a scat-back role to a bruiser role, so long-term receiving form is quite stable. The Depth Chart is admittedly a bit hacked, as it’s trained and tested separately only on 2017 data since I don’t have week-to-week depth chart data prior to 2017. Finally, age is significant and has been every week I’ve run the model.
As a reminder, to qualify for the model a player had to meet the following criteria:
- Minimum 5 opportunities per game (targets+rushes)
- Minimum 3 games played
Here’s the rest of season rankings heading into Week 7.
Rest of Season RB Ranks – PPR
Giovani Bernard ranks 31st in the model compared to Joe Mixon‘s 48th overall, but the experts have them flipped with Bernard at 64th and Mixon at 20th overall. If we just look at expected points, Mixon has Bernard covered by a 50.6 to 32.3 margin. However, Bernard has been much more efficient, racking up 46.7 PPR points (14.4 above expected) to Mixon’s 43.5 (7.1 below expected). It’s a matter of whether you believe the expected points, or the efficiency will win out. The model favors total production as the best predictor of future performance, which combines both volume and efficiency. Volume would have to drastically change for the experts to be more right than the model.
The model is probably too low on Ty Montgomery given he’s played injured at times. However, there is a chance now that Aaron Jones has put in some solid performances that he could eat into Montgomery’s workload going forward. Montgomery’s true rest of season value is likely somewhere between the expert rank and the model rank. Likewise for Jones — he should end up somewhere between 21st (model) and 37th (experts).
It doesn’t really matter whether it’s Rob Kelley or Samaje Perine as the early-down back — they both have extremely meh outlooks. Whichever of the two will garner more touches in a given week, they’re still only a back-end RB3 play per the model. Add in uncertainty and injury, and they’re both back end RB4-5s the rest of the way, which is closer to where the experts rank them. The guy you want is Chris Thompson.
Ameer Abdullah is RB32 rest of season per the model. He has just 65.6 expected PPR points through six games for a 10.9 per-game average. However, his efficiency has been putrid, with only 57 actual PPR points through those six games. Even if his efficiency rebounds to right at expectation, that still only puts him as RB27 per the model. Experts rank him as the RB18 the rest of the way, but that’s optimistic. He’d need both a volume and efficiency increase.
Both Mark Ingram and Alvin Kamara rank inside the top 17 per the model and per experts. Should one get injured, the other is almost surely an RB1 the rest of the way.