## SOE: Weekly Game Probabilities

Weekly game probabilities for week 2 are now up at Sports on Earth. Probabilities are a combination of the pre-season team strength estimates with a small dose of stats from week 1.

### 6 Responses to “SOE: Weekly Game Probabilities”

1. Anonymous says:

what were the probabilities for the Steelers-Ravens game?

2. Brian Burke says:

BAL 56%, PIT 44%
Score 23-20

3. Display Name says:

Can you add a summary table at the and of the article, of all your game probabilities, scores, and vegas line?

4. Anonymous says:

Hey Brian,

Since you are keeping track of how your model does against the spread, maybe you want to consider using something more scientific than your own whims for predicting the score based on the win probability the model spits out.

I've been using the data from http://wizardofodds.com/games/sports-betting/nfl/ to map from "estimated probability" of winning to fair "spread" (see table "Probability of Winning in the NFL by Point Spread"). That's my way of estimating what your model thinks a fair spread is.

Of course you could improve upon this. For example an 80% chance of winning between two offensively oriented teams should translate to a larger "fair spread" than the same 80% chance of winning between two defensively oriented teams.

Just some suggestions. I'm a huge fan.

5. Anonymous says:

For example, you are picking the Saints to win by 16 over the Browns, handily covering Vegas's 7 point spread. But your model only puts the Saints at a 65% chance of winning which, based on the data I linked to above, should translate to about a 4.5 point spread.

This means that I would pick the Browns to beat the spread based on the output of your model, but in your score predictions you are picking the Saints.

If you are just picking scores to have some fun that's totally fine, but I think it would be cool to track how well your model picks against the spread to the best of your ability.

6. Brian Burke says:

I'm not keeping track vs. the spread. SOE is. Ignore the scores. They are simply "plausible" scores given the game probability and respective offenses and defenses. There is absolutely no claim on my part to their accuracy.