Quote:
Originally Posted by ranchwest
Npt trying to be critical, but separating the surfaces seems critical to me. Turf and all-weather are nothing like dirt. It gets to be tricky because there's also track condition. And, of course, distance. And the factors. That's already a LOT of different models.
|
This first spreadsheet
is just dirt.
I wanted to get something that was functional up and running before I started breaking down it further like sprint/route, races loaded with speed/races with no speed etc.. And this is just graded stakes. The weights would be different for some other kinds of races.
Breaking it down is easy. That's just the selection criteria out of my database. That's a 10 second change to a query and running it. The tougher part is creating metrics that help pick more winners at greater value.
I want to add a consistency metric (working on it now) and I also want to play around with a running line selection criteria.
This spreadsheet is producing excellent results just looking at last race regardless of the conditions that day and today. So clearly I should be able to improve on this just by selectively looking at prior races and bringing in more relevant ones. But that's a lot of testing before I figure out the best way to do it. Then I can run it through the regression to see how to weight it vs. other factors. This is going to take me a long time, but when it's done it should be good. I already learned a couple of things and I'm just getting started.
