<
>

Follow-up: Prediction model goes 11-1 Sunday

10/5/2009

Posted by ESPN.com's Mike Sando

The chart shows Week 4 predictions from Brian Burke's model at Advanced NFL Stats.

Some questioned the 72 percent probability for the Broncos to beat the Cowboys, but Champ Bailey helped make the prediction come true. The Advanced NFL Stats model was "wrong" only on the Ravens-Patriots game. The model characterized the Bills-Dolphins game as a toss-up, while favoring the Vikings to beat the Packers in the pending Monday night game.

The model emerged from Sunday with an 11-1 record in games for which it projected a winner. As Burke wrote heading into Week 4:

For whatever reason, my model has been remarkably successful since its inception three seasons ago, and each year it has been slightly more accurate in predicting winners than looking at the consensus favorites. This is no small accomplishment, as luck plays a large part of many game outcomes.

Sometimes its predictions make you scratch your head, especially early in the season when most fans still carry notions of how good certain teams should be. It has its flaws, but it is brutally unbiased and hype-free. Of course, now that they are featured on nytimes.com, it’s due for an off year!

The model uses past game statistics to project how teams might fare. Gotta love those results.