## Game Probabilities - Week 15

Game probabilities for week 15 are available at the New York Times. This week I note the high playoff leverage of the NE-MIA game.

### 9 Responses to “Game Probabilities - Week 15”

1. Anonymous says:

Trying to reproduce these probabilities and would love some pointers.

The GWPs as given are 0.45 for Houston and 0.47 for Indy. Indy's had an average schedule (Opp GWP = 0.5) but Houston has had a weak one (Opp GWP = 0.45) and so I find that Houston's opponent-adjusted GWP is 0.45.

For the corresponding opponent-adjusted logits I get -0.565 for Houston and -0.120 for Indy. Adding on a penalty of -0.4619 for being away and computing to a win probability for Houston yields 29%, well under the 37% on the NY Times.

I'm using the Week 14 stats at http://www.advancednflstats.com/2013/12/team-efficiency-rankings-week-14.html and the -0.4619 figure is from http://www.advancednflstats.com/2011/09/weekly-game-probabilities.html

2. Anonymous says:

Do you use a standard 3-point (or certain percentage) adjustment for homefield advantage, or is it calculated in some other way?

3. Mike B says:

Houston's GWP is 0.41, not 0.45. And you don't have to do the strength of schedule adjustment. It is already factored into the GWP.

4. Anonymous says:

Thank you, Mike B.
I'm unsure of how to factor in home field advantage. The ahome coefficient at http://www.advancednflstats.com/2011/09/weekly-game-probabilities.html is not equal to the twice the constant term as http://www.advancednflstats.com/2009/01/how-model-works-detailed-example.html suggests it should be. The results I get by just ignoring the constant term and adding +/- 0.5*0.748577 to the logits for the home/away team gives results that don't match the NYT values but aren't far off; however, in this case I don't know what the constant term \$C['const'] = -0.4619 is even for.

5. Ben Cranny says:

What is the Packers probablilty of winning when you account for Rodgers being out?

6. Andrew Foland says:

I've backed out that the HFA adjustment for the NYT game probabilities is ~57.5% winning at home. This has actually changed from last season, when it was closer to 62%--I asked a couple weeks ago when that had been changed, but I think it was after everyone had moved to the next thread.

7. Anonymous says:

So @Andrew Foland, that would mean you add about 0.3022 to the logit for the home team and subtract it from the away team?

8. Brian Burke says:

Andrew-HFA has not changed. It does not directly correspond to a probability. When teams are close in strength, HFA is larger than when the game is lopsided. It averages 57.5%.

9. James says:

The equation is: (H*(1-A)*F) / (H*(1-A)*F+(1-H)*A*(1-F))

Where H is the GWP of the home team, A the GWP of the away team, and F the home field advantage constant of 0.575. I've used this to get the game win probabilities a day early for the whole year and they are always within rounding error.

### Leave a Reply

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