RB Projections Experiment - Week 10 with Results

[Challie]

I've been playing with some formulas to predict RB performance. Though I'm still toying around with it, here's what would have been predicted for Week 10:

Rk Name Team Opp Predict Results
1 Shaun Alexander SEA STL 13.4 7.0
2 Tiki Barber NYG ARI 12.0 14.0
3 Curtis Martin NYJ BAL 11.8 18.2
4 Derrick Blaylock KC NO 11.8 17.2
5 Jerome Bettis PIT CLE 11.8 17.5
6 Ahman Green GB MIN 11.6 7.6
7 Chris Brown TEN CHI 11.3 -0.6
8 Clinton Portis WAS CIN 11.3 5.6
9 Kevan Barlow SF CAR 11.2 15.0
10 Jamal Lewis BAL NYJ 11.1 4.6
11 Deuce McAllister NO KC 11.0 12.4
12 Edgerrin James IND HOU 10.9 5.0
13 Domanick Davis HOU IND 10.9 19.6
14 Emmitt Smith ARI NYG 10.6 15.2
15 Corey Dillon NE BUF 10.6 7.8
16 Michael Pittman TB ATL 10.4 3.9
17 Rudi Johnson CIN WAS 10.3 9.1
18 Fred Taylor JAC DET 9.8 8.3
19 Anthony Thomas CHI TEN 9.8 4.1
20 Warrick Dunn ATL TB 9.7 4.1
21 Marshall Faulk STL SEA 9.5 6.0
22 Eddie George DAL PHI 9.4 7.7
23 Brian Westbrook PHI DAL 9.3 11.9
24 Willis McGahee BUF NE 9.0 2.4
25 T.J. Duckett ATL TB 8.2 14.5
26 Lee Suggs CLE PIT 7.5 -0.1
27 William Green CLE PIT 7.3 0.0
28 Kevin Jones DET JAX 7.3 4.4
29 Onterrio Smith MIN GB 7.3 7.7
30 Brad Hoover CAR SF 7.0 0.8
31 Nick Goings CAR SF 6.7 4.6
32 Troy Hambrick ARI NYG 6.7 1.1
33 Dorsey Levens PHI DAL 6.6 9.5
34 Najeh Davenport GB MIN 6.4 0.3
35 Moe Williams MIN GB 6.3 8.7
36 Shawn Bryson DET JAX 6.2 0.7
37 Steven Jackson STL SEA 6.1 8.8
38 Michael Bennett MIN GB 6.1 1.1
39 Tony Fisher GB MIN 5.9 9.3
40 Maurice Hicks SF CAR 5.8 2.9

While it's far from perfect, the goofy formula I've been using would have had C-Mart, Blaylock and Bettis higher than I would have & the formula would have been a good thing. The only high-ranked dud was Chris Brown and Alexander, to an extent. Also need to keep in mind that in my numbers I had Staley forecast as sitting out, so Bettis' prediction was more reliable than fantasy sites who predict on Wed or Thu.

Let me know if anybody is interested in me trying this again for Week 11.

[matmat]

interesting.

what scoring system was this for? the numbers look a little suspect?

I've been toying around with something like this, but the models I've been thinking about have been a little too complicated for me to just sit down and write them down.

[Challie]

The model is a little bizarre, it's too boring to list the complete formula, but mainly

Three numbers are averaged together for each RB: 1) His average performance during the season, 2) His team's average production from the RB spot - weighted against his likely role in the given week*, 3) The opponents average yield to RBs.

* e.g. with Priest out, Blaylock goes from being predicted as 10% of the RB offense to being 90% of the RB part of the offense.

The numbers are more based on a 10yds per point yardage league with TDs being 6 points.

Like I said, it's far from being perfect - haven't even begun to try to factor in turnovers which seems pointless. And all RBBC guys suffer in the predictions.

I wouldn't put much stock in this, but I'll run it for a few weeks and see if it can develop any consistency. At the very least, it might provide some food for thought. But it's not very sophisticated (there's no allowance for a defense getting better or worse or being injured). I'll post a week 11 in a few minutes. Almost done with that. At worst, feel free to ignore it.

[matmat]

i guess I was trying to figure out why the result from blaylock wasn't something like 26, but rather 17.... almost as if you counted 20yds/point

[Challie]

Damn, you're right about the results. I took my week's results from one of my Yahoo leagues where I have 20yds/pt. I'm fried - I'm in too many leagues. A frustrating 6-4 in most of them.

Hey, I'm trying week 11 here: http://www.fantasyfootballcafe.com/foru ... p?t=113252

Obviously, don't bet the house on any of this.