Calling Week 1
This week I projected Clemson to win by a touchdown over 11th ranked LSU. In Baton Rouge. Most of my picks aren’t this spicy. This is the first game of the season for both teams, so most of the data feeding my model is from last season. I am not gambling, and apart from the fact that I just generally don’t bet, this is why.
I’d wanted to do this for years but never made the time. A couple of years ago I was between jobs, so I finally found the data and built something. I work with data for a living, so it was an easy way to stay sharp on something interesting. So now every Monday through the fall, I pull team stats for the AP top 25 teams and make a prediction of that week’s outcomes.
Last year my success rate was 74% on the season. The favored team wins straight up roughly 75% of the time on average. So, I created a very elaborate way to pick the favorite. The last week of the season I did manage to correctly call 90% of the games. There are some gimmes at the end of the season so a lot of other people did well, too. Still, I’m pleased with this.
Survive and Advance
During the off season I figured out how to forecast scores. This will be the real test for my model. My model compares one team’s offensive stats with their opponent’s defense (and vice versa). This ignores a lot. Like:
- rosters
- strength of schedule
- bye weeks
- ???
I hope to be competitive, but the stakes are low. It turns out that I like keeping up with a few games each Saturday to see how I did. I’ll text a few friends to tell them if I think their team will win. And I am always keenly interested in my model’s projection for Texas.
Some have asked if I’m betting. I’m not. I won’t. Nothing is riding on a bad call. But if I’m right then I get to taunt you.
Hook’em!
Photo by Daniel Foster, CC BY-SA 2.0, via Wikimedia Commons