This is what my data tells me.
Most important for the data you probably have:
1. Morning Line
2. High-level factor such as BRIS Prime Power. (We use HDW's PSR.)
3. Speed Ratings, time-weighted. (i.e. newer ratings are worth more.)
4. Trainer standings at the track.
5. Early running style.
6. Jockey standings at the track.
We use similar to these in our software. Biggest difference is that we have 3 different high-level factors (i.e. like BRIS Prime Power).
Thus, ours looks like:
1. ML
2. PSR
3. Older version of PSR.
4. RTG (A proprietary factor that using 32 other factors weighted.)
5. Speed Ratings (weighted for races back).
6. Trainer at the track.
Once these numbers are computed, they must be normalized to 100%.
Steps to finalize
1. The final percents must be skewed towards the top by about 9%. That is, build a "normal" for the race. Example, 10% in a 10-horse field is a 1.00. In an 8-horse field 12.5% becomes 1.00.
2. Raise all of the "IVs" to the power of 1.09.
3. Normalize once again to 100%.
This will give you some pretty good pool percentages which can then be turned into odds.
|