Week 6 Estimated Win Probabilities

[TDs3nOut]

I had never seen this until this morning, but it looks as if the NYT might run a piece each week in which Brian Burke of Advanced NFL Stats publishes his estimated probabilities of teams winning their games. Here is that piece this week:

http://www.nytimes.com/2013/10/10/sports/football/week-6-nfl-game-probabilities.html?_r=0

A couple of his estimates that caught my attention are that .39 is his estimated probability of both NE winning at home against NO and Buffalo winning at home against Cincinnati. Both of those estimates seem low to me. Interestingly, it seems as if they both put too much emphasis on the Bengals having beaten the Pats last week. Was that game reliable evidence for extrapolating into the future, or was it mere random noise?

 

[boltfan72]

Wow, there are a lot of crap numbers in there.

Baltimore, at home, has only a 30% chance against GB?
Houston's probability of winning is only 9 points lower than Denver's?
Being the favorite in the game with the largest point spread in NFL history only gives you a 92% chance of winning? Does anyone in their right mind think Jax would win once in 10 tries, in Denver?
Patriots at 39% at home? Wow.

 

[Fountain City Blues]

Wow, there are a lot of crap numbers in there.

Baltimore, at home, has only a 30% chance against GB?
Houston's probability of winning is only 9 points lower than Denver's?
Being the favorite in the game with the largest point spread in NFL history only gives you a 92% chance of winning? Does anyone in their right mind think Jax would win once in 10 tries, in Denver?
Patriots at 39% at home? Wow.

I don't think it is that crazy to think, there is more parity in the Nfl than most think. That, and NE is simply unimpressive right now.

 

[TDs3nOut]

Wow, there are a lot of crap numbers in there.

Baltimore, at home, has only a 30% chance against GB?
Houston's probability of winning is only 9 points lower than Denver's?
Being the favorite in the game with the largest point spread in NFL history only gives you a 92% chance of winning? Does anyone in their right mind think Jax would win once in 10 tries, in Denver?
Patriots at 39% at home? Wow.

Those estimates do seem questionable. Your interpretation of the estimate that Denver wins is interesting. Perhaps that highlights why estimated probability of winning hasn't caught on as a popularly reported statistic. The idea of football games played as repeated trials of an experiment doesn't seem particularly intuitive.

 

[MilkSpiller22]

Wow, there are a lot of crap numbers in there.

Baltimore, at home, has only a 30% chance against GB?
Houston's probability of winning is only 9 points lower than Denver's?
Being the favorite in the game with the largest point spread in NFL history only gives you a 92% chance of winning? Does anyone in their right mind think Jax would win once in 10 tries, in Denver?
Patriots at 39% at home? Wow.

Not sure how he came up with these numbers... wish that was explained more, wish he gave us formulas... Definitely interesting, but just like the spread, nothing means anything...

 

[TDs3nOut]

Not sure how he came up with these numbers... wish that was explained more, wish he gave us formulas... Definitely interesting, but just like the spread, nothing means anything...

The article does mention that the probabilities are estimated using a statistical model. I suspect that the model uses data on things like number of points teams have scored and given up, which team is at home, and perhaps the betting spread on a game in order to estimate the probability that a team wins.