These were the previous 2 days ratings which take account of a vast array of different things. What the model does is takes the racecards and my personal HRB ratings for the day, loads them all up and compares them with the vast database I've now compiled, and checks for all kinds of possible features of importance in a deterministic regressor before presenting the output. So for Flat Handicaps run today, it decided that these were the main important factors. I create my own race z-scores and ELO ratings for horses, jockeys, trainers, stallions and damstallions, so you'd have to create your own, and if you know HRB you'll see it has a stat attack feature and I get it to calculate the different stat attacks myself but it also considers things like more recent form, percentage of runners beaten (and squared), and previous odds / pedigree. You can see from this model that for these types of race today it places great weight on the last HRB score achieved.
You'll also see how often my top rated horse is the market leader, but in the list of determining factors below, the model is completely blind to today's actual odds.
So the previous two days are attached, the ratings are rudimentary and converted to resemble Timeform figures as opposed to being on anything like their scale, but even so, it's achieved a 37% strike rate and a 22% return on investment. Where horses haven't won, it hasn't had to go too far down the list most often to find the success. I'll try and post tomorrow's sheet before going off to work.
The HTML is still being worked on but I am developing it to provide commentary on each horse's chances. When done it will rate the horses accurately based on their chances on the day, and give strong evidence for why a horse is being backed.
