[antonyp22]

I've created an AFL model which comes up with a projected score for each team based on data from previous rounds in the same season.

The model goes at 80% ATS when the difference between the predicted margin and opening line is > 5 points.

I've verified it and made sure I'm not using data from round 5 to predict round 5 scores e.t.c.

It still seems to good to be true, is there any other way of testing the model?

[mjespoz]

antonyp22,

Try testing your model on at least a couple of complete seasons of games out of time. See how you go then.
Testing your model on the same data that you used to build it can very easily lead to overfitting.

Cheers mate.

[antonyp22]

Thanks for the tip mjespoz!

The model is built on data from the current season so betting would only commence in round 5 after the first 4 rounds have been completed (to gain a base set of data). So I'm not sure if it would be applicable on games out of time.

[EXhoosier10]

1) get 2012 data
2) run the same process using first 4 round data
3) try to predict round 5

rinse and repeat for other seasons

edit: reread and you may have already done this