Quote:
Originally Posted by Jeff P
I have been told at least some of the whale teams are paying observers for physicality information.
And yes, they do have quantifiable algorithmic methods for incorporating physicality observations.
Interesting sidebar:
I have been told the observers are compartmentalized.
All communication is supposed to be one way. The observer's job to simply submit physicality observations and nothing else.
The system is designed so that the observers have no way of knowing who the other observers are, how to contact team members they don't have a need to know, or whether or not their physicality observations are actually used in the team's fundamental model.
This is done to make it difficult for team members and observers to go out and form a new team on their own while taking other team members, source code, and/or the team's fundamental model, etc. with them.
Which apparently, has actually happened a few times.
-jp
.
|
This is interesting, though I'm skeptical to what extent they can presently (and reliably) integrate qualitative paddock observations with their fundamental models.
First, there would probably be significant variability in terms of how individual observers rate horses across the physicality dimensions.
Second, you can't buy historical datasets so you need to build these observational datasets ground-up which would take time.
Third, I'm not convinced that the "average" signal from any one of these physicality dimensions is particularly powerful as behavior can vary quite dramatically across horses such that even with a large sample the average impact of a particular dimension may be compressed.
Anecdotally, I've tried (albeit on a small-scale) to manually gather some of the typical physicality info for dozens of races at my local track and then incorporate them into my model. They just don't seem to pop from a significance standpoint. Obviously, it could be that I'm just not a good paddock-person.
