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.
Game Probabilities - Week 15
published on 12/12/2013
Subscribe to:
Post Comments (Atom)
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.
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
Do you use a standard 3-point (or certain percentage) adjustment for homefield advantage, or is it calculated in some other way?
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.
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.
What is the Packers probablilty of winning when you account for Rodgers being out?
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.
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?
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%.
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.