Football is still 10 long days away. Rodger is out for the week, except for daily pre-packaged posts (presumably, with less nutritional value than the fresh ones). I've already read the latest post from Bring Your Champions, They're Our Meat, twice. I even found myself reading an Iowa blog, looking for something to pretend to be outraged about. And still, my Northwestern football needs have not been satisfied.
How about some baseless speculation from a website with questionable methodology?
Before finding them mentioned on that not-to-be-named Iowa blog, I had never heard of WhatIfSports.com. From my 15 minutes of Wikipedia and browsing their website, they seem to have started out as a nifty idea-- using the overwhelming amount of sports data available to simulate games between historical teams that never played each other-- and then grew and went corporate, selling their statistical souls to Rupert Murdoch and Microsoft.
In this adventure, however, they aren't trying to pit teams from the past against each other. Instead, they're running simulations of 2010 teams, presumably predicting the final result. The predictions business is tricky, though, since eventually, those teams will actually play each other and, uh, prove you wrong. For instance, again from the Iowans, WhatIfSports' simulator made a bold prediction for the 2010 Orange Bowl: Iowa 80, Georgia Tech 17.
Now, that was just one simulation run by some guy on the internet. For their 2010 Big Ten predictions:
Using our college football simulation engine, Whatifsports.com simulated thousands of college football games to create this Big Ten Preview. Team ratings, player ratings and depth charts are accurate as of July 31st. What you see in the WIS-Big Ten Predictions table is the most-likely outcome based on the computer simulations. The team-by-team schedules use Absolute Records, which you can learn more about below. Those same game-by-game results also generate average points per game for both teams.
So, run thousands of game simulations and use those to determine probabilities.
For the politically inclined, it's the same basic idea as what Nate Silver at FiveThirtyEight does. Except where Nate Silver built an incredibly sophisticated model that uses constantly updating polling numbers (adjusted based on pollster error rates), augmented by fundraising data, demographics, and a host of other pieces of predictive data, the folks at WhatIfSports use... ?
That part's not so clear, actually. The paragraph quoted above is the closest I found of any explanation in methodology. This is different from their historical simulations since a lot of the players-- new starters, freshmen, etc.-- don't have much by way of statistics behind them to feed into the simulator. Never mind that it also has no way of accounting for new coaches (see Vanderbilt), injuries, or any number of other factors.
I mean, midseason predictions give you something. You've got half a season of useful information to work off, and maybe some common opponents, so you can test the transitive property of football. Preseason predictions give you nothing useful except what we already knew. It's a replay of last year, minus Mike Kafka, plus some dude's imagined idea of what Dan Persa and Venric Mark might be like. But I've got a feeling that that guy probably hasn't been camped out in Kenosha watching him practice. It's all got to be absolute guesswork.
I'm saying, take this with several grains of salt. The predictive value here is very low.
And since it probably does rely so heavily on last year's statistics, it's no surprise that its predicted season is, uh, a lot like last year. Screen shot of results:
So, to recap: A simulator of little to no predictive value and unknown methodology says that the same Big Ten teams that were good last year will be roughly just as good this year. Which you probably already knew. I mean, it's not like Coach Fitz is going to change anything he was already doing based on these predictions.
But that's not inflammatory enough. How about:
SCIENCE PROVES NORTHWESTERN WILL BEAT VANDERBILT!
That was sort of fun, wasn't it? Anyone? Maybe if I added hip hop references...