## Game Probabilities Week 8

Game probabilities for week 8 are up at the New York Times. This week I take a look at how lopsided this weekend's matchups are.

In terms of probabilities for the favorite and underdog, a game is typically about a 62-38 percentage affair. When the underdog wins almost 4 out of 10 games, it can make for an exciting Sunday afternoon. But this Sunday probably won’t be one of them...

On the other hand, when there are so many mismatches, there is bound to be a big upset. Statistically, it’s very unlikely that all those favorites will win...

### 14 Responses to “Game Probabilities Week 8”

1. John Black says:

Translation: "this might happen, that might happen, blah blah blah". I seen better analysis on Key and Peele las night.

2. Anonymous says:

Is Jaguars-49ers game calculated as a neutral field game? (game is in London)

3. Brian Burke says:

Yes. Neutral site was factored in.

John Black-You probably don't even know that K&P is supposed to be funny.

4. Anonymous says:

SF over JAX is "only" a 64 percent split, not 70 as it says in your column.

5. Rod says:

..and what would be the approximate percentage of at least two underdogs pulling an upset this week?

Thanks a lot, your work is awesome.

6. Matt says:

Is there a way to use this probabilities to help choose the best bets against the spread?

7. Anonymous says:

Seattle vs St Loius - with Sam Bradford out and Kellen Clemens in should not be an 89% chance of winning. It should be 95%. Only if Russel Wilson gets hurt can St. Loius win, with Kellen Clemens as they QB

8. Anonymous says:

Another typo in your article. In paragraph two it should be (62 minus 38) not (68 minus 38).

9. Anonymous says:

Matt, the way I've been doing it is by converting between the spread and the moneyline. You can easily find some conversions online. The one problem is that there often aren't moneylines available for big spread games (it's harder for the books to balance the sizes of the bets in those cases) so I've guestimated what the moneyline for those large spreads should be. The difference between Brian's percentages and those implied by the spread would give you the confidence you should have in a pick (according to Brian's model anyways). I haven't tracked this as diligently as I could have but in the small sample I've tried, Brian's model has done extremely well on games where there's larger than a 15-20% win probability difference. It doesn't seem like there are any games like that this week (in fact, the discrepancies as a whole are much smaller this week; hopefully this is a one week fluke and not some systematic adjustment) as the "most confident" pick this week is Cowboys +3 (Brian's model at 51%; the implied win% from the spread is 41%).

10. EpicWestern says:

Matt - The problem with betting the spread is that there's a price attached to it. A prediction telling you that a team will beat the spread doesn't tell you if that team will beat both the spread and the price.

With moneyline bets however, if a team is a betting underdog, and its line is positive (its not just a slight underdog), predicting that team will win (over 50% probability) implies that a bet on it will be good. So far Brian's picks have been crushing in that department.

11. Mitch says:

Yes you can convert the efficiency ratings into a spread, example Det would be -.5 over Dallas.

I,ll post a few of the biggers diff in the team rankings week 7 thread. I've posted those there the past few weeks and the model has yet to lose.

12. Mitch says:

According to the model Seattle -15 over the Rams with Bradford no less.

The model strongly suggest Seattle would be a strong play this week.

13. Mitch says:

The model has SF -14 VS Jax.

SF is due a huge regression and playing in London where they probally don't want to be will more then likely just go through the motions in this game, knowing it's a easy win.

The model is suggesting a small lean to Jax plus the points.

14. Unknown says:

Mitch, how exactly are you converting the efficiency ratings into a spread? I was looking at a correlation between the probability and the ML but would be interested to see how get your numbers.

### Leave a Reply

Note: Only a member of this blog may post a comment.