Here's another phenomenon in models and factors I've noticed on occasion.
If you test a factor on "all" cases you will typically find that you'll lose the track take or thereabouts and tend to discard it.
I've seen cases where a factor seems to have no value at very short prices (everyone else sees it too) and very long prices (the horse may have a positive going for it but it's simply not good enough to win in today's race) but have some value among contenders that are not big favorites.
So maybe a guy like Benter incorporated the live odds into his model, but it may make sense to incorporate odds ranges also.
__________________
"Unlearning is the highest form of learning"
|