Quote:
Originally Posted by MJC922
Horse against horse using an iterative process. Lengths, distance, surface, weight. The computer estimates the quality of the effort based upon the finish, that's 99% of what's being used in a nutshell. Data processing and no judgment.
|
Your ratings are similar to my own
automated ratings, but I only make them for stakes races because there are so many complicated class categories nationally it would become an overwhelming job to do it for everything. It sounds like you found a way to deal with that problem.
My original goal was only to test Class Ratings vs. Speed Figures and settle that internal struggle I was having for decades anyway.
Ultimately, it became obvious to me I could generate automated comparative class ratings that were very close to as good as speed figures (1% lower win% over 6+ years), but I also learned that when I combined them, the results were superior to either alone (about 2%-3% better than either alone).
My own automated ratings tend to suffer when impressive 2yo MSW winners of unknown quality are moving into stakes, ALW horses of unknown quality are trying stakes against older, a horse wins by a huge margin and it's hard to tell in an automated system whether he was very good, the field very bad, or a little of both etc..
In my gambling, I resolve these issues with a more subjective analysis of the field and race in a similar way to how Timeform Europe makes their numbers. But I think there are still situations where the speed/pace figures scream the right answer and it's not so clear with a comparative class analysis. The problem of course is that the speed/pace figures generally tend to drag you towards the favorite and the goal is to find horses that are better than they look on paper so you have a chance to get a better price.