Thursday, September 23, 2010

WVU @ LSU: The AccuScore Simulation

Well, take this for what its worth, but West Virginia has already lost to LSU... approximately 6,300 times.

AccuScore, an internet-based statistics think-tank, makes its bones by pumping their statistical model full of stats from around the NCAA and the real-world, then applying that model to each week's slate of games 10,000 times each, predicting the probability of each teams victory based on the elements of the model and the results of each set of simulations. Their NCAA simulations were 57.6% accurate against the spread last week... everyone on this site is that good, and none of us use math.

"Shfifty-five"?

The brain-farts at AccuScore produced 10,000 simulations and gave the LSU Tigers a 64% of victory at home against the Mighty Mountaineers of West Virginia U Saturday night. What their models cannot account for, and this is the inherent failure of all statistical models, is that at the moment the ball is put into play in the highly-anticipated Big East-SEC match-up Saturday night, the model is obsolete because new data is now available that has not been applied to the model. The old "you-can-throw-the-stats-out-when-these-two-teams-get-together" dynamic.

But more there is more to this model that just doesn't add up. A closer look at the simulated stats shows that the teams will be relatively even in the turnover battle (1.6 for WVU, 1.8 for LSU), yet LSU has a better TD projection despite WVU's superior -- albeit marginally -- passing, receiving, AND rushing stats (see below):
WVU
Pass Rtg - 126.6
Pass Yds - 181.5
Pass TD - 0.9
Pass INT - 0.5

Rush Yds - 122.8
Rush YPC - 4.0
Rush TD - 0.7

Rec Yds - 156.6
Rec YPC - 12.0
Rec TD - 0.8

LSU
Pass Rtg - 122.0
Pass Yds - 158.4
Pass TD - 1.3
Pass INT - 0.8

Rush Yds - 113.1
Rush YPC - 3.6
Rush TD - 1.0

Rec Yds - 122.7
Rec YPC - 12.3
Rec TD - 1.0
No, this is not a spreadsheet of Brandon Hogan's blood-alcohol contents, but the statistical averages across the series of simulations. So there you have it: LSU 22.8 - WVU 16.4. In the immortal words of Mr. Lee Corso, "NOT SO FAST MY FRIEND!"

How LSU is projected to be almost an entire TD better than WVU, based upon those other offense statistics is really beyond me. Factor in that Jordan Jefferson may actually be clinically retarded, and I am just not seeing this.

The model also projects Tavon Austin to have just 3 reception and ZERO carries for the entire game. Jeff Mullen (who is a toss-up with Jefferson in Quiz Bowl), will surely have some wrinkles in the game plan to get the budding superstar more than 3 touches in this game.

Did I forget to mention that Noel Devine is playing with his Heisman hopes on the line? Yeah, he'll probably have a little something to say Saturday night too.

Expect the Mountaineers D to stand tall (with or without Hogan), and the offense to continue to unleash the fury of speed and talent that it showcased during the 1st half of the Maryland game.

Take the points America, and in my humble opinion, take the Mountaineers STRAIGHT UP!

- Grey Bush

AccuScore Game Forecast Preview [AccuScore.com]
AccuScore Game Forecast [AccuScore.com]

No comments:

Post a Comment