NFL Model

[BigJohnny]

I've created a model to predict lines, using 2008-2010 statistics.

From CBS sports take the current power rankings, and subtract the difference between the two teams. Multiply by a factor of .5.

IE NE = 1, BUF=32 difference is 31x.5 = 15.5

+/- 3 points for home field

@NE = NE expected to win by 18.5
@BUF = NE expected to win by 12.5

From weeks 3-12 HFA worth +/- 2
From week 12-end of reg. season HFA worth +/- 3
Playoffs HFA worth +/- 4.5

If your expected value is more than 3 points either way, IE NE expected to win by 18.5, and the odds are at -15, I lay 1 unit on the Pats. Same vice-versa.

I backtracked to 2008 and came out with about ~65% winning percentage.

Good luck fellas.

[Indecent]

I've created a model to predict lines, using 2008-2010 statistics.

From CBS sports take the current power rankings, and subtract the difference between the two teams. Multiply by a factor of .5.

IE NE = 1, BUF=32 difference is 31x.5 = 15.5

+/- 3 points for home field

@NE = NE expected to win by 18.5
@BUF = NE expected to win by 12.5

From weeks 3-12 HFA worth +/- 2
From week 12-end of reg. season HFA worth +/- 3
Playoffs HFA worth +/- 4.5

If your expected value is more than 3 points either way, IE NE expected to win by 18.5, and the odds are at -15, I lay 1 unit on the Pats. Same vice-versa.

I backtracked to 2008 and came out with about ~65% winning percentage.

Good luck fellas.

Which years did you use to develop the "model"? How did you decide to use +-2 for 3-12, etc?

From what you have described I would be concerned that you are data-mining.

[hugh4310]

I like this, good luck, and keep posting

[TheBet]

Interesting concept, surprisingly, using this method, the lines come pretty close to the actual lines (back tracking two weeks). Not sure how it holds up long-term.

As for HFA, http://www.twominutewarning.com/hfa.htm seems to support a hypothesis that HFA is worth more as the season progresses.

[Indecent]

As for HFA, http://www.twominutewarning.com/hfa.htm seems to support a hypothesis that HFA is worth more as the season progresses.

Yep, read that article. Didn't have it saved though, thanks for the reminder.

In regards to my post, I was more curious as to whether those were the first values for HFA he used or whether there was some trial and error involved in picking the parameters.

[TheBet]

I applied this model using ESPN power rankings and SBR odds to weeks 3-12 in NFL 2010.

Came out with 59 'plays'. Using the closing lines from SBR odds, I ended with 67% winning ATS (40-19).

Sample size too small to conclude anything, but from the looks of it, this 'model' is a winner for the 2010 season.

[BigJohnny]

My model indicates that tomorrows play is Arizona +7.

Good Luck!

[bringdownthehous]

My model indicates that tomorrows play is Arizona +7. Good Luck!

I understand what your trying to do here, and it is an interesting concept, but wouldn't taking ARZ tomorrow be questionable considering dating back from 08/09 they had warner, boldin, etc.

This is a completely different offensive team from past years, and stats taken from those years with warner would be skewed?

Idk maybe im thrown off by the whole concept, but yeah im a dallas backer tmrw night nd wanted some info on why u like arz

[LLXC]

From what you have described I would be concerned that you are data-mining.

Data mining is okay; overfitting is not.

[TheBet]

I understand what your trying to do here, and it is an interesting concept, but wouldn't taking ARZ tomorrow be questionable considering dating back from 08/09 they had warner, boldin, etc. This is a completely different offensive team from past years, and stats taken from those years with warner would be skewed? Idk maybe im thrown off by the whole concept, but yeah im a dallas backer tmrw night nd wanted some info on why u like arz

I interpreted it as taking the weeks current power rankings, as ESPN adjusts their power rankings before each weeks game. So it wouldn't take into account how the Cardinals did in 08.

Once again, this is a good concept. It allows major sports 'statisticians' from ESPN/CBS to do the leg work for you. The concepts are basic, but with minor tweaks here and there, this could be a serviceable model.

[Indecent]

delete

[Indecent]

delete

[FreeFall]

you could also do a page rank one the the two teams scores and then get your teams ranked that way as well.

[BigJohnny]

BigJohnny's play looking good thus far!

[BigJohnny]

The model is 1-0 in real play.

[BigJohnny]

BigJohnny's 12/26 NFL Plays:

Miami -3.5
San Fran +2.5
Baltimore -3.5
Oakland +3