[EXhoosier10]

If I have a model that is estimating a team's goals for a game, how can I create a corresponding moneyline for that game?

Say I am estimating these teams with their corresponding goals for tonight's game (not real numbers, just a hypothetical scenario):

Kings - estimated goals of 2.5
Ducks - estimated goals of 2.75

I'm not sure how I can related these numbers and turn them into probabilities of the ducks beating the Kings. Ya, the Ducks are -135, but with this information, should I be bet on the Ducks because they are likely to win more than 57% of the time?

My only thoughts are trying to find out what the probability is of the ducks scoring this many goals if the kings will actually score 2.5 goals and then say that they have a xx% chance of winning if that is the case (however, I still can't figure out a way to come up with that xx%).

Any help is appreciated,

Kevin

[maxdalury]

use the two variable poisson calculator. hockey is the best fit for the poisson distribution out of all the major sports.

[EXhoosier10]

max, if both events have the same expected values, that doesn't give me a -100 odds.

I'll have to look into it more tomorrow, but any more help would be great

[20Four7]

If I have a model that is estimating a team's goals for a game, how can I create a corresponding moneyline for that game?

Say I am estimating these teams with their corresponding goals for tonight's game (not real numbers, just a hypothetical scenario):

Kings - estimated goals of 2.5
Ducks - estimated goals of 2.75

I'm not sure how I can related these numbers and turn them into probabilities of the ducks beating the Kings. Ya, the Ducks are -135, but with this information, should I be bet on the Ducks because they are likely to win more than 57% of the time?

My only thoughts are trying to find out what the probability is of the ducks scoring this many goals if the kings will actually score 2.5 goals and then say that they have a xx% chance of winning if that is the case (however, I still can't figure out a way to come up with that xx%).

Any help is appreciated,

Kevin

If they are likely to win 57% of the time that translates into a money line of about -133. If your being offered -135 it's not a good bet. It comes down to how good are your % of winning.... if that becomes 60% the line should be -150 and it becomes a good bet.... if ti's really 55% then the dog is a good bet because it should be around +125.

[EXhoosier10]

If they are likely to win 57% of the time that translates into a money line of about -133. If your being offered -135 it's not a good bet. It comes down to how good are your % of winning.... if that becomes 60% the line should be -150 and it becomes a good bet.... if ti's really 55% then the dog is a good bet because it should be around +125.

I guess that was poor writing on my part. I understand converting lines to their corresponding break-even win%.

What I can't figure out is how to create my own win % based on a teams expected goals vs. an opponent. If I could do that, then I can see if there is value in a given line.

[BrewBrain]

I guess that was poor writing on my part. I understand converting lines to their corresponding break-even win%.

What I can't figure out is how to create my own win % based on a teams expected goals vs. an opponent. If I could do that, then I can see if there is value in a given line.

The only way I think would work is to run your goal prediction process against games already played and see what percentage of the time favorites/dogs at each predicted goals differential win, then apply that percentage against tonight's lines.

I did this for the NBA - created a formula for determining my own custom spread, applied the formula against last year's games and am using the winning percentages to wager on this year's games.

For instance, 1-point favorites according to my custom spreads won about 55% of the time (give or take) last year. When I calculate a 1-point spread on a game this year, I compare my favorite/dog percent chance of winning (55%/45%) against the money line offered and bet whichever offers value, if any.

In theory, I think this should work. In practice, my winning percentages this year so far are quite a bit different from last year, so I'm losing. Short season so far though.

Sample size is also an issue - I had at most around 150 or so games at the most frequent spreads calculated. Several seasons of data is probably needed.

[TomG]

You can calculate a team's Pythagorean Expectation (and thus Win% and ML) using the inputs of 1) goals scored 2) goals allowed.

http://en.wikipedia.org/wiki/Pythagorean_expectation

I'm not sure which exponent is best for NHL. Starting out, you should try experimenting with several exponents and see which ones give you the most stable result. Or Google hunt to see if others have done similar research for NHL.

You bigger concern should be in the accuracy of your estimate of each team's expected goals.

[maxdalury]

thats because its giving you the three way outcome so there is a chance of a tie... so you could halve the chance of the tie and add it to each side to get the appropriate win probabilities or use some more advanced calculation of the chances of a particular team winning the shootout.