Last week we published twelve numbers. Ten were off, a little too high. Projected David Montgomery for 77 yards, he went for 10 rushing yards. We assumed the carries from week 1 would matter, welp they did not. This miss led to finding a bug in the model, which is good news. This means we can fine tune and move forward.

How our Week 1 projections actually did

Want to audit last week, look no further than this table here.

RB Team Proj Actual Miss Wk1 car Wk2 car Car change
Jahmyr Gibbs DET 112 52 -60 29 16 -13
Kenneth Walker III KC 98 117 +19 23 24 +1
Derrick Henry BAL 97 68 -29 24 16 -8
Ashton Jeanty LV 85 48 -37 23 21 -2
Bijan Robinson ATL 85 72 -13 21 16 -5
Breece Hall NYJ 84 29 -55 22 16 -6
David Montgomery HOU 77 10 -67 20 6 -14
Jonathan Taylor IND 77 92 +15 19 24 +5
D’Andre Swift CHI 76 45 -31 18 16 -2
Cam Skattebo NYG 73 36 -37 18 12 -6
Jacory Croskey-Merritt WAS 66 43 -23 16 12 -4
Rhamondre Stevenson NE 65 43 -22 18 6 -12

Typical miss: 34 yards. We said ±26. We over-projected 10 of 12, by an average of 28 yards.

Ten of the twelve running backs had a decrease in carries. Across all twelve, 251 carries in week 1 to 185 carries in week 2. That figures to be a ~26% decrease.

Where the miss came from

Source of the miss Yards
Carries that did not repeat -255
Yards per carry being off -85
Total -340

It appears ~75% was volume which accounts for the miss.

Workload regresses too

The chart below shows a really interesting finding on where a running back’s carries go after a big volume week.
Notice how every section, starting from fewest carries, leads to a bigger fall.
What’s that saying, the bigger they are the harder they something.

Change in a running back's carries the following week, by workload

Carries this week n Avg this wk Avg next wk Change % who lose carries
5-9 1,415 7.0 8.4 +1.4 42%
10-14 1,280 11.9 12.2 +0.3 49%
15-17 572 15.9 14.0 -1.9 62%
18-20 391 18.9 14.5 -4.3 76%
21-24 269 22.1 16.1 -6.0 83%
25+ 152 27.4 17.8 -9.7 91%

The rule: next week’s carries ≈ 5.5 + 0.50 × this week's, against a league average of 12.7.

Carries this week Expect next week
10 10.5
15 13.0
20 15.5
25 18.0
30 20.5

Half of any given week’s workload is gone by the next week, and vice-versa.

“It’s just blowouts” … not so fast!

That week’s game n Avg carries Next week Change
Won by 14+ 75 24.8 17.1 -7.7
Won by 1-13 177 25.1 16.7 -8.4
Lost 73 24.3 17.5 -6.8

You may be thinking, of course there will be a ton of carries in a blowout win, but I caution against thinking this. According to the table above the change is bound to happen in either direction whether a team won or lost. Actually, maybe the thinking is pretty close to correct.

So is workload useless? Not quite, it’s still the best thing we have

Do we ignore workload/carries? No, they are still very stable over the long haul.

Metric Correlation, same back, consecutive weeks
Carries +0.357
Rushing yards +0.237
Yards per carry +0.048

So we fixed it

Projecting Week 3 yards from Typical miss r
Week 1 carries only 27.3 +0.214
Week 2 carries only 26.7 +0.274
Weeks 1-2 average 26.4 +0.290

The model already shrinks running backs over expected, but was not shrinking carries per game yet. Truth it, we did just miss this. Was not on the radar.

High-workload backs (≥18 car/g) Bias Typical miss
Week 1, before the fix -17.4 32.0
Week 2, before the fix -21.0 33.9
Week 3, before the fix +3.1 30.6
Week 6+, before the fix +2.7 32.5

That bling spot can now see, let the light in!

Model Typical miss Weeks 1-2 Bias on high-workload backs
What we published last week 26.2 25.9 -18.8
+ shrink the workload 26.4 26.5 -1.8
+ last season’s workload 26.3 26.1 -2.6

The Week 3 traps

Be wary of a running back who comes out of nowhere with a huge workload on any given week, they might let you down the next game.

RB Team Wk1 car Wk2 car 2-wk avg What we believe The read
Emanuel Wilson SEA 2 21 11.5 11.5 the waiver add we are being told to make
Aaron Jones MIN 12 23 17.5 13.5
Omarion Hampton LAC 12 23 17.5 13.5
De’Von Achane MIA 11 20 15.5 12.9
Bucky Irving TB 8 17 12.5 11.9
James Cook BUF 13 21 17.0 13.4 led Week 2 in rushing yards
TreVeyon Henderson NE 0 16 16.0 12.4 brand-new role, so some inferring
Jonathan Taylor IND 19 24 21.5 14.9 two straight — the real workload
Kenneth Walker III KC 23 24 23.5 15.5 two straight — the real workload

Your intuition is correct after all, a workhorse is a workhorse. Stick with them as their carries will remain and be consistent. That speculative one, let them remain speculative.

Week 3 projections

Typical miss on these is ±26 yards.

RB Team Wk 3 opp Car/g YOE/g Opp D Proj yds Proj TD Read
Kenneth Walker III KC MIA 23.5 +46.5 +49 83 0.56 under-scored the chances
Derrick Henry BAL DAL 20.0 +23.6 +48 77 0.71
Jahmyr Gibbs DET NYJ 22.5 +10.9 +10 72 0.63
Jonathan Taylor IND HOU 21.5 +6.6 -69 71 0.76 TD luck this week
James Cook BUF LAC 17.0 +23.5 +2 68 0.53 hot week → projected off carries
Ashton Jeanty LV NO 22.0 -18.3 -33 68 0.53 cold week → yards should return
D’Andre Swift CHI PHI 17.0 +14.0 -5 64 0.61 cold week → yards should return
Breece Hall NYJ DET 19.0 -18.2 +4 64 0.50 cold week → yards should return
Bijan Robinson ATL GB 18.5 -8.0 -13 60 0.44
Kyren Williams LA DEN 11.5 +16.4 -57 60 0.46 hot week → projected off carries
Chase Brown CIN PIT 18.0 -8.7 -8 60 0.50
Tony Pollard TEN NYG 10.5 +4.1 +112 60 0.37
Omarion Hampton LAC BUF 17.5 -9.7 +67 59 0.50
Christian McCaffrey SF ARI 10.0 +7.3 +60 58 0.55
Bucky Irving TB MIN 12.5 +9.9 -35 55 0.44

Let us see what tomorrow brings