NFL 7pt teasers: -3.5 to 3.5

[djiddish98]

Intuitively, one would think that crossing the 3 twice would be about the best thing you could do in an NFL teaser, even if you have to waste points crossing the 0.

Yes, there is likely some "push decay" in crossing that second 3 (from +2.5 to +3.5), since we are farther from the true line. However, I think (looking at some data) that crossing that first -3 from -3.5 is about the most valuable move you can make. It makes up for the "push decay" on the second 3.

However, looking again at historical spread data shows that if I teased every -3.5 to +3.5, I would not cover enough to give an edge on a 7pt teaser.

Why is this? Is my data bad? I've noticed that -3.5's have under performed a good deal against the spread. Should they really be -3, and hence, we lose the value of crossing that first 3? Anything else?

 

[MrMonkey]

I'm not qualified to discuss this matter but if some qualified posters can that would be great! Always wondered around the -+3 number and what to do in buying .5 points? Sounds like dog at +3.5 is better in your data?

 

[djiddish98]

Correct.

In my data (from sunshine forecast; thanks SBR), -3.5 Faves covered the spread 45.36% of the time (out of 571 observations), and -3.5 faves won by exactly 3 11.03% of the time, a huge push %.

This leads me to believe that -3.5 faves are really more like -3 faves in disguise. The next step would lead to the thought that blindly betting +3.5 dogs would be profitable, but I'm inclined to think that this is just a pattern from the past and don't want to bet money that it will continue this year.

For example -3.5 covering the spread by year:

2009 - 47.62% (21 Obs)
2008 - 30.77% (13 obs)
2007 - 52.63% (19 obs)
2006 - 55% (20 obs)
2005 - 61.11% (18 obs)
2004 - 45% (20 obs)

and so on. I often curse the NFL for having such small sample sizes.

Also, I hacked all this together yesterday, so I'm not 100% sure that the math is correct.

 

[djiddish98]

Also, does anyone believe that the home dog bias has disappeared? I feel like you need many years to fully evaluate these things, but in looking at 2000 - 2009, I see away faves actually covering the spread 50.5% of the time, as opposed to 1978 - 1999, they covered the spread 45.51% of the time.

Again, with the small samples, I'm tempted to bring in as much data as possible, but I also don't want to pollute my good data with bad data. And I would think that data from the 2000's would be better than data from the 80s.

 

[rito]

I've heard that sunshine data isn't real accurate.

Surely the home dog bias is gone the market is too smart for that now.

That's what you need to remember when checking on past performance of teasers. Home wong dogs hit like 78% for years, but were also hitting 55%ats, you need to assume going forward that will drop to 50% and evaluate the teaser from there.

 

[djiddish98]

Thanks for the input, durito. I was thinking something should be done to correct some of the noise in the past data, as opposed to just taking the historical tease win percentages at face value.

Any opinion on "push decay", if that term is even correct - should it be factored in or is it not that significant?

 

[djiddish98]

I also have 3 years of pinny data I pulled from SBR, although I feel like this is too small a sample to gauge things with.

The nice thing about that data is that it also has the odds on the spread, so I can know if a spread is leaning one way or another. However, I'd have to do some adjustments to convert all their shaded 9's and 1's into spreads that are closer to 50/50.