Quote:
Originally Posted by classhandicapper
I have one question.
Let's assume you rate a winner 15 off his last race.
He comes back 2 weeks later and disappoints relative to that figure. Over the next couple of weeks 3 more horses come out of the same race and disappoint relative to their figures.
Do you go back and adjust that race weaker?
If yes, the next time that original winner comes back will he show a different figure than a 15 for that race?
That's the one thing I don't have in my automated system that I may like to add.
My reports show me the Par for the class and a Class Rating for the race based on who was actually in it so I can identify strong and weak fields for a class.
I've tested the strong/weak process. It has a couple of quirks, but it's pretty good.
What I don't have is an automated follow up process to see how horses are coming back so I can refine the ratings further as I have more information. I do that manually when I'm handicapping. I'd like to automate something there eventually and see how it impacts the results of my tests.
Do you do that?
|
Yes I do that, well not me but the computer does that several hundred times per night actually. It's what I mean about iteration hundreds of times, the figure may have started out as a 15 but by the time the dust settles it may be more or less than that based upon the latest results. It may not even be results from the winner either, it could be let's say an also ran that came back and ran well or ran poorly. For a period of time everyone gets a 'say' in what the past race should be. Maybe the also ran came back and ran well to win and the horse he beat in that race then came out of the subsequent race to run even better, now we're two races away and were actually getting 'opinions' from horses who were not even in the original race as to how it currently rates. It's a lot of data processing. To do it manually you'd be limited to thinking in basic terms of maybe one full iteration and then the follow-up would be pretty much confined to just the horses exiting the race, it would quickly become impossible to tie it all in manually so hundreds of horses are giving input. I mean you can still get good numbers from one or two 'limited' iterations I guess you could call them, that's kind of what most speed handicappers are doing.
The other thing is I don't use class pars aside from that one recent change I mentioned for firsters where I 'seeded' a par number for the firster because there is no previous information available. A field of 100% firsters then is a crude classification for the computer to make. That's definitely still the weakest scenario for accuracy, but the way that works now is about as good as I could get it without bringing times into mix and there aren't enough of those races for me to even put that on the to-do list tbh, it's a limited accuracy problem that takes care of itself pretty quickly. I can live with the three times per year people want to point out how far off the computer was.
I think that's where people generally start their assumption for how a class rating would be calculated though, it sounds like that's kind of what you're doing currently, start with a class par. The process I'm following though does not take that path, the differences between the horses are all coming from their races as they run against one another globally in a sense. You could extract a class par out of that when it all finishes iterating but it doesn't begin with a class par.
