Kenneth. Walker. III. +74 yards over expected, what a game.
What does this mean for week 2? Will he repeat his performance? Or will regression happen. Let’s find out.
How we grade a week
Not so fast! That fireworks show of a week 1 for offenses was fun, but is it significant going into next week? We declare it is not.
Every running back carry since 2020 gets an expected gain from where it started: yard line, down, distance, score, and clock. Then, fitted by ordinary least squares on the five seasons before this one. The expectation explains almost nothing about any single carry (rushing is mostly noise for that), however, it is calibrated and it adds up: over a game it says what a back’s carries were worth. A back’s week is then:
- Yards vs expected
- TDs vs expected
Expected here refers to opportunity (it’s rating), not what’s expected of the running backs themselves.
Week 1: who beat their carries
| RB | Team | Opp | Car | Yds | xYds | Yds vs exp | TD | xTD | TD vs exp |
|---|---|---|---|---|---|---|---|---|---|
| Kenneth Walker III | KC | DEN | 23 | 173 | 99 | +74 | 1 | 0.7 | +0.3 |
| Kyle Monangai | CHI | CAR | 10 | 100 | 44 | +56 | 1 | 0.2 | +0.8 |
| D’Andre Swift | CHI | CAR | 18 | 124 | 70 | +54 | 3 | 1.4 | +1.6 |
| Derrick Henry | BAL | IND | 24 | 144 | 99 | +45 | 3 | 1.2 | +1.8 |
| Jahmyr Gibbs | DET | NO | 29 | 156 | 114 | +42 | 2 | 2.2 | -0.2 |
| Christian McCaffrey | SF | LA | 10 | 68 | 44 | +24 | 0 | 0.1 | -0.1 |
| Jonathan Taylor | IND | BAL | 19 | 98 | 75 | +23 | 2 | 1.3 | +0.7 |
| Saquon Barkley | PHI | WAS | 15 | 83 | 70 | +13 | 0 | 0.1 | -0.1 |
Who fell short
| RB | Team | Opp | Car | Yds | xYds | Yds vs exp | TD | xTD | TD vs exp |
|---|---|---|---|---|---|---|---|---|---|
| Rhamondre Stevenson | NE | SEA | 18 | 51 | 84 | -33 | 0 | 0.1 | -0.1 |
| MarShawn Lloyd | GB | MIN | 13 | 37 | 59 | -22 | 0 | 0.2 | -0.2 |
| Rico Dowdle | PIT | ATL | 8 | 15 | 35 | -20 | 0 | 0.1 | -0.1 |
| David Montgomery | HOU | BUF | 20 | 60 | 79 | -19 | 2 | 1.4 | +0.6 |
| Quinshon Judkins | CLE | JAX | 12 | 33 | 51 | -18 | 0 | 0.4 | -0.4 |
| Bijan Robinson | ATL | PIT | 21 | 83 | 97 | -14 | 0 | 0.1 | -0.1 |
| De’Von Achane | MIA | LV | 11 | 36 | 49 | -13 | 0 | 0.1 | -0.1 |
What one week is worth
| Predicting next week’s rush yards from | Typical miss (yds) | R² |
|---|---|---|
| Guess the average | 29.7 | 0.00 |
| This week’s yards | 28.1 | 0.09 |
| This week’s expected yards | 27.7 | 0.12 |
| Workload + season-to-date beat + opponent | 26.2 | 0.19 |
- A back’s beat in one week has no relationship to his beat the next. - Across 363 Week 1 → Week 2 pairs the correlation is 0.00.
- The carries predict next week better than the yards do. - Expected yards beat actual yards as a predictor, because the carries are the part the coach controls.
- Over a season, the beat becomes real. - it will come more into play in a few weeks, just less so now.
*author’s note: “Guess the average” is just the baseline. For every back, predict that next week’s rushing yards will be the league-average RB total from the training seasons (about 47 yards, for backs with ≥5 carries the week before), regardless of who he is or what he just did.
What Week 2 should look like
Typical miss on these is ±26 yards
| RB | Team | Wk 2 opp | Car/g | YOE/g | Opp D | Proj yds | Proj TD | Read |
|---|---|---|---|---|---|---|---|---|
| Jahmyr Gibbs | DET | BUF | 29.0 | +42.4 | +67 | 112 | 1.12 | hot week -> projected off carries |
| Kenneth Walker III | KC | IND | 23.0 | +74.3 | -27 | 98 | 0.71 | hot week -> projected off carries |
| Derrick Henry | BAL | NO | 24.0 | +45.2 | -33 | 97 | 0.91 | hot week -> projected off carries |
| Ashton Jeanty | LV | LAC | 23.0 | -1.7 | +2 | 85 | 0.60 | |
| Bijan Robinson | ATL | CAR | 21.0 | -14.1 | +29 | 85 | 0.54 | |
| Breece Hall | NYJ | GB | 22.0 | +7.8 | -13 | 84 | 0.73 | |
| David Montgomery | HOU | CIN | 20.0 | -19.0 | +93 | 77 | 0.80 | |
| Jonathan Taylor | IND | KC | 19.0 | +23.1 | -29 | 77 | 0.76 | |
| D’Andre Swift | CHI | MIN | 18.0 | +54.0 | -35 | 76 | 0.81 | hot week -> projected off carries |
| Cam Skattebo | NYG | LA | 18.0 | +7.1 | -22 | 73 | 0.64 | |
| Jacory Croskey-Merritt | WAS | DAL | 16.0 | -5.8 | +48 | 66 | 0.54 | |
| Rhamondre Stevenson | NE | PIT | 18.0 | -33.1 | -8 | 65 | 0.46 | cold week -> yards should return |
Caveats
- ±26 yards. The model explains a fifth of week-to-week variation. The rest is game script, injuries, and bounces. Or sometimes coaches change their mind, happens.
- Rushing only. Receiving work is not in these numbers, which doesn’t do players like Bijan Robinson justice.
- Role assumed constant. A back who lost carries to injury or a benching in Week 1 is projected on that reduced role.
- Opponent defence after one game is nearly noise, which is why the model treats it how it does. Last season’s number carries more.
- One season of Week 1 -> Week 2 history per year. Six seasons of data shows us.
Are we saying week 1 doesn’t matter? Of course not, it’s just not the most important thing to consider when trying to predict how these running backs will do next week.