I checked the half point calculator for the push probability of 0 when the spread is 1 for regular season NFL. The half point calculator shows a probability of 0% for the draw. We know a draw is possible in the regular season. Is the probability so low that the half point calculator rounds it down to zero, or the half point calculator assumes a draw is not possible? Thanks for your help.
-
-
It is a very rare event. I would probably use the push rate for all games where abs(spread)<=7, and use that as an equal value for all lower spread games (where spread <9).Comment -
This website lists the likelihood of a draw at 0.2%: http://www.marasoft.com/nfl-margin-victory.html
It's only happened 19 times since 1974 when the sudden death OT was installed in the rules. That's 35 years * 256 regular season games per year = 8960 games. 19/8960 = ~0.21%
I assume it's just registering a very small number and rounding to zero.Comment -
The half point calculator uses 1.5 point radii to estimate push probabilities. So, only ties with Spreads within 1.5 points would be included in that data. Perhaps it's never happened and perhaps it's rounding down. In any case, it's a poor tool to use for the data you're looking to obtain. You would need your own database to perform any type of Bayesian inference regarding ties.Comment -
Obviously the probability of a tie is not zero for ANY spread, and would probably be greater for Spreads approaching pick.
Basically, this just manifests the error (0.2% in this case) of using only a range of spreads to determine a push frequency, instead of more rigorous methods that would reduce the error further.
At a minimum I would use 0.25% for pick and downgrade to 0.15% for 10 or higher, so that the integral would approx. equal 0.2%.Comment -
Say you have a stale line -110 on +3 that gives you a "4% edge". Is the possibility of a push reflected in this 4%? Or does it need to be tweaked down a bit to reflect the potential push?Comment -
Since you don't win/lose on a tie, the push frequency is not part of the EV edge calculation.
Using your example, assuming NFL push frequency of 10% for spread of 3, with no edge your pick would have 12/21 - 1/10 = 47.14% chance of covering and 9/21 = 42.86% chance of losing, and 10% chance of tie. Since you have a 4% edge, that implies your pick now has 48.86% chance of winning, 41.14% chance of losing, and 10% chance of tie.
Your EV is (0.4886 - 0.4114*11/10)/.9 = 0.04 = 4%.Comment
