Tommy,

I think you are on the right track.

The best way to test your odds line is to keep track of how well it performs. One fairly easy way to do this it to group horses into different pct chance of winning "buckets." (For those of you that are math inclined I am suggesting that Tommy create a histogram.) For example, group all the horses that your model predicts will win in the 25% to 35% range and then compare how often they actually win to the predicted probability. If in the above example the horses that you predicted would have a 25% to 35% chance of winning actually do win about 30% of the time, you can start feeling pretty good about your odds line (assuming you have lots of data). You can start feeling even better if you see similar results for all the other "buckets" (e.g., 0%-5%,5%-15%,15%-25%, etc.)

Of course, such an accurate model does not guarantee success. I could build a "system" that converts the post-time odds into a pct chance of winning and it would prove to be a very accurate predictor of a horse's true odds of winning (despite popular opinion the betting public as a whole is pretty darn good). Unfortunately, I would lose at a rate just about exactly equal to the takeout. So you need to have an accurate model that can point out overlays and underlays to turn a profit (in other words an odds line that is as accurate as the betting public's but also has some frequent, contrary and correct opinions).

It is easy to be contrary and incorrect. It is also easy to go along with the crowd and be correct about 1/3 of the time. The challenge is to be contrary and correct often enough to show a consistent long-term profit. Building an odds line is a good way to start down that latter path.

Good luck with your model. Even if it never quite pans out you will learn a lot about handicapping and value in the process.

Chris