## Game Probabilities - Week 4

This year the weekly game probabilities are featured on the nytimes.com Fifth Down. Each week, I'll post a link to the probabilities at Fifth Down.

The model has been updated this year to add the 2007 and 2008 seasons. Previously, it was based on data from the 2002-2006 seasons.

The only significant change is that I have re-included defensive interceptions this year. I had based the decision to exclude them on the lack of auto-correlation for team defensive interception rate from the first half of the season to the second half in both 2006 and 2007. However, the 2008 season indicated a relatively strong auto-correlation. In short, I based my previous conclusion on too small a sample. Ultimately, I adjusted the model weight of defensive interceptions by how well it predicts itself throughout the season on average in those three seasons.

### 14 Responses to “Game Probabilities - Week 4”

1. William T says:

Nice job Brian. When I relaunch my site I'm going to incorporate your win probability calculations, based on your formula and my data set.

2. John Candido says:

Congratulations Brian. That's a great opportunity for you as well as pioneers the way for the rest of us doing similar statistics. I have one question regarding the resampling? What was your new overall prediction rate? I know that you hold 75% to be the highest possible given luck, based on your past article. Was the prediction correct still up there? I have been experimenting taking out Week 1,2 and 17 on my model to get a more realistic appraisal. Again I am running up against 75% at best.

3. Brendan Scolari says:

Good stuff. One question (and this is probably dumb), what is auto-correlation?

4. Brian Burke says:

I'm sorry. It means how well a variable correlates with itself over time.

What I did was divide the season in half and tested the correlation of each efficiency stat with itself between the two halves of the season. This is a rough measure of how well a stat predicts itself. Some stats are relatively consistent, and some are more random.

5. Sam W. says:

Congratulations on your success, Brian. My only worry is now that you're on NYT.com that all the guys in my suicide pool will now have found the same pot of gold that I found several years ago. You're a genius in my books. Thanks for sharing all your hard work with us "not-so-good-at-numbers" types.

6. Anonymous says:

good stuff right here

7. Mr.Ceraldi says:

Hey Brian;
Some numbers from your Game Prob. don't make intuitive sense to me
GB 8% >Min in GWP and 7% greater in SOS and is
16% behind MIn in your Prob.(58% to 42%)
Buf is only 5% > in GWP to Mia and 13% LESS in SOS
yet they are 50% to 50% in your game probability?

What am I missing?

Shouldn't a stronger team with stronger schedule recieved stronger game prob.
By your efficency and SOS table GB is closer to Min then Buf is to Mia?

BTW my formula (using your numbers) has GB and
MIN closer 47 to 53% and Mia Buf 55% to 45 Mia??
thanks
Dan

8. Brian Burke says:

Mr. C-Thanks and good catch. Yes, I did not apply opponent adjustments to the game probabilities. They are baked into the GWP here, but the game probs at NY Times do not have past opponent strength adjustments. Next week I'll have the correct probs. This year I've switched over to a completely automated system for doing these calculations, and this is the test week.
Here are the game probs with past opp strength factored in:

0.22 Detroit at Chicago 0.78
0.63 Cincinnati at Cleveland 0.37
0.1 Seattle at Indianapolis 0.9
0.63 NY Giants at Kansas City 0.37
0.48 Baltimore at New England 0.52
0.32 Tampa Bay at Washington 0.68
0.45 Tennessee at Jacksonville 0.55
0.31 Oakland at Houston 0.69
0.29 NY Jets at New Orleans 0.71
0.37 Buffalo at Miami 0.63
0.24 St. Louis at San Francisco 0.76
0.37 Dallas at Denver 0.63
0.42 San Diego at Pittsburgh 0.58
0.5 Green Bay at Minnesota 0.5

9. J.P. says:

congrats on your model....you are perfect for WEEK 4 even including tonights MNF game since your probability calls for 50/50
keep up the great work....I love seeing probability play itself out right
CHEERS

10. Anonymous says:

Overall, I am pretty impressed.

jim in Georgia

11. Anonymous says:

Where do I find week 5 probabilities?

12. Anonymous says:

I do ther same with NCAA College Football Picks every FRIDAY.

Does anybody mind if I post them here?

Jim
Stockbridge, GA

13. Anonymous says:

Jim you do this with college games? Where do you post these predictions currently?
thanks

14. Anonymous says:

Hi Brian- Your site is amazing, and I love your weekly game probabilities. I'm in a unique pool that I've been trying to attack mathematically.

Quick description: Pick 3 underdogs to win outright. If they win, you get the spread as points for the week. There's a weekly winner and cumulative points winners for the season. Since it's all about accumulating points, it doesn't make sense to go for 1 point games, or three 2 point games. There's strategy in picking high, medium, and low.

My thought is to use your probabilities as expected value, with the spread as the multiplier. In this way, one could balance the probability of the underdog winning with the points that come with the win. Given the probabilities, however, I think the high spread games would always be selected using this method.

Any ideas on how to normalize this to get decent long term results?

-Craig