Clearly the track surface was changing through the course of the day. But each time a decision is made to change the variant you decrease the sample size. And if the track is really changing, even between just one or two races, and the variant is similarly adjusted, then the numbers get less and less accurate because the error rate goes up (normal consequence of reducing the sample size). But there isn't any other way to deal with it, since there is no way to collect more data (more races) for the given conditions. My only suggestion then would be to do something similar to what the animal breeders (cattle) do. That is, when they assign breeding values for sires (called EPD's, for Expected Progeny Differences), they also include an Accuracy number, expressed as a percentage, that reflects the sample size that the EPD value is based on. In other words, for a ten race card, a variant that is constant through the day (or multiple days?) gets say 10/10, or 1.00. If the variant for a race is based on fewer races, say 4/10, then the accuracy stat would be 0.40. Then the handicapper can make the decision as to how much faith to put into the number. The numbers-maker will have done his level best to provide an accurate number, given the total information (however sparce) available to him.



Post Edited (06-10-04 03:00)