More Trakus problems, this time SA

[GBL]

Originally Posted by Robert Fischer

Hi GBL,

Those are two models that could occasionally be significant.

Ground loss is the easy/obvious one... You're basically starting by flagging All ground loss..., and then filtering out only the instances where the ground loss happened in places or scenarios, where the ground loss actually matters in races.

The Race Shape, pattern recognition stuff is a little more complicated, and requires a competent programmer, probably with some related experience(in data patterns), who can follow my instructions and translate that into an application/model.

Hi Robert,

Thanks for the reply.

I understand from your message that your emphasis probably is on 'Pattern Handicapping'.

For pattern handicapping to work, various frameworks and matrices would need to be defined and within each framework, and corresponding matrix, relative factors, angles, variables, and other dynamics (both tangible and intangible) would needed to be assigned and weighted based on its impact and importance. Obviously some of the frameworks and matrices works best for one level of racing than the other.

[Robert Fischer]

Originally Posted by GBL

Hi Robert,

Thanks for the reply.

I understand from your message that your emphasis probably is on 'Pattern Handicapping'.

For pattern handicapping to work, various frameworks and matrices would need to be defined and within each framework, and corresponding matrix, relative factors, angles, variables, and other dynamics (both tangible and intangible) would needed to be assigned and weighted based on its impact and importance. Obviously some of the frameworks and matrices works best for one level of racing than the other.

Sounds complicated.

I'll have to re-read all that, once I have some coffee.



For starters, you do the simple, significant stuff.


TimeformUS already does a nice job with what should be the obvious first step (basic Race-Flow, into three classes [neutral, forward-favoring, collapse].

You do race flow. And if you want to get fancy, you create a cousin, incorporating odds into the flow model.


Simple significant stuff for the Positioning. Lone Speed? Duel?


Simple significant stuff for Path Bias. Paths 4-6+?, Paths 1-3?

Want to get fancy, create a cousin for Path Bias, and incorporate whether or not a horse was physically positioned inside or outside of other horses.




There's other stuff that takes more trial and error, and may be silly/pointless, but is nonetheless interesting.

Optimum Positioning. (in a neutral or soft pace, the Optimum Positioning would be on the lead. In a hot pace the Optimum Position would be somewhere off the lead, and it could be 'fudged' using the aforementioned race flow data).

Lots of other similar 'idea' model possibilities... But the meat and potatoes of whole point is the few simple models that are known to be significant.

[46zilzal]

Originally Posted by davew

During the running of the race some tracks have little chicklets showing each horse position. Are these from Trakus information? Occasionally a horse is not shown or looks like it is manually adjusted as it makes it to current position.

yes but since they are coming from about where the horse's HIPS are, it has some error

[46zilzal]

Originally Posted by Dave Schwartz

CJ,

Trakus uses a satellite system, right?

If so, it seems logical that a local system would be more accurate.

It NEVER was local. The Trakus tech told me that initially the signal went all the way to BOSTON to be "processed" and then back to the source (in my experience, Woodbine) before it was available to be then fed into the local database.