Early Experiment in QB Formula (key word EXPERIMENT)

[Demented Avenger]

On average, passing attempts are negatively correlated with wins, so I'm not sure that simply multiplying pypa * expected number of attempts is the way to go. Why not just use average yards per game, which is the same stat with just one variable to predict instead of two?

[Challie]

Here's a wrap up of how it went:

First Last vs. Opp Diff Actual Rat
Drew Bledsoe vs. NYG -13.26 6.34 19.60
Matt Hasselbeck vs. HOU -5.98 9.36 15.34
Peyton Manning vs. STL 2.00 16.12 14.12
Carson Palmer vs. TEN 3.77 17.64 13.87
Kerry Collins vs. SDG -9.59 3.84 13.43
Ben Roethlisberger vs. JAC DNP 11.80
Mark Brunell vs. KAN 9.20 20.72 11.52
Drew Brees vs. OAK -1.95 9.03 10.98
Jake Plummer vs. NEW 6.57 17.54 10.97
Eli Manning vs. DAL -3.97 6.80 10.77
Tom Brady vs. DEN 2.01 12.58 10.57
Michael Vick vs. NOR -1.76 8.79 10.55
Anthony Wright vs. CLE -1.27 8.21 9.48
Steve McNair vs. CIN -6.83 1.53 8.36
Trent Green vs. WAS 4.84 12.87 8.03
David Carr vs. SEA 2.23 10.18 7.95
Brian Griese vs. MIA 0.99 8.45 7.46
Trent Dilfer vs. BAL -10.51 -3.06 7.45
Joey Harrington vs. CAR -8.31 -1.38 6.93
Jake Delhomme vs. DET 4.45 10.97 6.52
Kyle Orton vs. MIN 4.70 10.54 5.84
Aaron Brooks vs. ATL 10.87 16.58 5.71
Byron Leftwich vs. PIT 2.88 7.59 4.71
Daunte Culpepper vs. CHI -3.02 1.44 4.46
Kelly Holcomb vs. NYJ 7.15 11.34 4.19
Marc Bulger vs. IND 2.63 6.42 3.79
Vinny Testaverde vs. BUF -0.39 3.27 3.66
Gus Frerotte vs. TAM 1.51 1.79 0.28

A few surprises, I'm going to keep tinkering. Things worked out fairly well.

Considering the comments above about how QB points are not completely predictable, I should think that's obvious. So many factors, luck among them. This is not an attempt so tell you with certainty that Bledsoe will get 17 points, or Peyton will get 14. It just attempts to show what the numbers of past performance can try to predict & give you some clues where you might not have realized things before ... much like PLindsey's Def picks. I've got a long way to go towards working this out, but having looked at this for a while, I'm encouraged.

The stat that turned out to be pretty good was injuries - they were based solely on players listed in the injury report and their status. (By the way, my earlier post didn't include offensive injuries in the numbers for predictions, these ones do.)

I'll post a Week 7 attempt and see how this goes. Hopefully it might provoke some thought or discussion.

Have a great week.

[Plindsey88]

You need to factor in injury status somehow. For instance, Roethlisberger may not play (at least that's the last I heard). Overall, I think you're playing with fire when hoping to predict human production/behavior from a formula. Plindsay's defensive rankings can be pretty accurate, but you're factoring in a team's worth of players. Overall, that accounts for a big game from one individual and a poor game from another. Overall, it all averages out. However, when you're trying to predict how an individual is going to do based on a formula, I think you're in for some bad headaches! You can't factor in all the variables that could affect production -- injuries, last-minute sicknesses, an unexpected death in the family, etc. While it's nice for conversation's sake, I wonder just how accurate you'll ever really be with something like this??

I agree.... Been there - tried this.... It is doable for defenses because there are so many players involved.... One guy having an off night doesn't kill the projection.... But it does not work consistently for individuals, because there are just too many possible variables.... Did the guy get laid last night, is he fighting with his wife, did he eat a good breakfast, is he sore from practice, are his receivers suffering from the dropsies today, etc... etc... etc...

Often times the slightest little thing can impact a guys performance.... Remember in week one when Peyton missed a wide open Harrison for about a 50 yard TD? It rattled him... It affected his performance the rest of that game.... Things like that are impossible to predict..... On defense if your starting safety tweaks an ankle, it's rare for that to kill the entire performance of the rest of a top defense.... Typically top defenses have more going for them than one guy.... But if you're trying to predict the outcome of one player, and he tweaks an ankle, your whole formula is for not....

Believe me, I've tried.... Good luck with this, but you will find that in the long run your guesses will be just as accurate as your formula results.... If not more so....

[DigitalGambler]

Yep good luck... I had some of the most convoluted formulas in the world last year... sometimes right, sometimes wrong... I gave it up this year and found that I'm not spending as much time pouring over the stats and I'm having more fun.