[million2one]

I created a model for the NBA that gives me the expected outcome of the game.
Assuming a normal distribution, I know the difference of the two team scores is the mean outcome, but what data do I use to calculate the variance and thus the standard deviation.

Thanks for any help

[million2one]

Do I calculate the average absolute deviation in my model and use this as the standard deviation??

[HUY]

I created a model for the NBA that gives me the expected outcome of the game.
Assuming a normal distribution, I know the difference of the two team scores is the mean outcome, but what data do I use to calculate the variance and thus the standard deviation.

Thanks for any help

Your model is specifying the mean so the model should specify the variance as well. If you let us know of the internals of your model then we might be able to help.

[matthew919]

If you're modeling each team's score separately, you might think of each team's result as a normal distribution (it's really more of a binomial, but you can use a normal to approximate it if n is large enough).

Then you can use the fact that the sum/difference of two normal distributions is also normal, with variance equal to the sum of the two variances. So compute the variance for each team, and just add them up.

But as HUY says, without knowing more about your specific approach it's tough to help.

[million2one]

Thank you both.
Your comments have been helpful and I think I figured it out now.

[Capper007]

Did you apply what you learned?

[million2one]

I am constantly applying what I learn, problem is I don't have enough time these days to track everything.
How do I know if my model is valid?
Should I compare my line to the opening or closing vegas line, or should I be comparing my model to the actual outcome of the game?

[Waterstpub87]

I am constantly applying what I learn, problem is I don't have enough time these days to track everything.
How do I know if my model is valid?
Should I compare my line to the opening or closing vegas line, or should I be comparing my model to the actual outcome of the game?

All of the above. Comparing your model to the outcome of games will provide an measure of overall accuracy. Comparing against the Vegas line will give the market opinion for the closer. The easier thing you can do is measure the error.

Put your prediction, the closing line and the actual result in three columns next to each other in excel. Take the error of your prediction against the result, and then the line against the result. Square the difference of each individual game, and then add them together. Divide this by your number of your sample points and take the square root. Do the same for the closer. If your number is lower, you are doing alright.

[million2one]

excellent advice, thank you.