Predicting Speed Rating from Past Ratings

[TrifectaMike]

Originally Posted by Grits

An honest question because I get the feeling your story is contrived. Have you forgotten what you've stated to others, here, in the past... that you are (or were) the educator of grad students at a major NY university? While now, you present yourself as new to handicapping? This is one past thread, summer of 2011. I admit, I'm woefully lost.

http://www.paceadvantage.com/forum/s...2&page=1&pp=15

Just a simple tale. Seeking a solution to question A and discovering something more important.

Mike

[sjk]

It argues for a better way of making speed ratings. There ought to be information in the back races.

[classhandicapper]

>>Perhaps there is a cycle of equine performance consistency. W3 could be the minimum performance consistency point.<<

<Bingo!!!

Mike<

I don't buy this at all (especially because it is not consistent with controlled data I had someone else evaluate).

The 3rd race back will randomly sometimes be a peak, sometimes a trough, and sometimes something in between.

It will sometimes randomly be first off a layoff, 2nd off a payoff, 3rd off a layoff or 10 off a layoff.

It will sometimes randomly be the horses 1st start in a horse's career or his 100th start.

There are unquestionably form cycles, certain races a horse is more likely to run back to than others, and some that should weigh more than others, but you aren't going to determine that the 3rd race back is less significant than the 4th race back with a huge pile of uncontrolled data (which is what you seem to be implying, but correct me if I am wrong).

I think the result you got is related to the data not being controlled for all the factors that impact figures. If you apply it broadly, it may outperform doing nothing at all, but it will lead you astray quite often.

[Tom]

Originally Posted by sjk

It argues for a better way of making speed ratings. There ought to be information in the back races.

As opposed to basing them on reality?

[Robert Fischer]

I could be way off base, but a couple/few concepts come to mind here.

1. First of all, I think the example was simplified a bit to illustrate the methodology. (In other words, the example itself is not necessarily the whole point of the exercise.)

2. TrifectaMike's 'Bayesian hierarchical modeling' looks really powerful/interesting.

3. Some of the issues brought up by posters questioning the example, 'exemplify' a belief of mine that the most powerful results are often produced by a marriage of both insight and methodology.

[Greyfox]

Originally Posted by Robert Fischer

2. TrifectaMike's 'Bayesian hierarchical modeling' looks really powerful/interesting.

What is powerful and interesting about it?

In fact, does anyone have a clue here as to what "Bayesian hierarchical modeling" is, and if so show us how it is done and show us how it improves on any other methodology with an example predicting a race on tomorrows card at any track?

[Tom]

1. First of all, I think the example was simplified a bit to illustrate the methodology. (In other words, the example itself is not necessarily the whole point of the exercise.)

hat example?
How is any of this related to handicapping or racing?

[ReplayRandall]

Originally Posted by Tom

hat example?
How is any of this related to handicapping or racing?

Gave this a read.......looking for an answer to how and why this has correlation to a horse racing methodology.........yes, maybe, no??

http://en.wikipedia.org/wiki/Bayesia...hical_modeling

[Greyfox]

-

Originally Posted by ReplayRandall

Gave this a read.......looking for an answer to how and why this has correlation to a horse racing methodology.........yes, maybe, no??

http://en.wikipedia.org/wiki/Bayesia...hical_modeling

We're not stupid and know that Bayes is a mathematical approach to certain decision making situations.

Show us how you can get a better speed figure using it.

[whodoyoulike]

Originally Posted by whodoyoulike

Were you the grad student? So, how do you predict speed ratings from past ratings?

Originally Posted by TrifectaMike

With a Hierarchical Bayes Model

Mike

Thanks. But, I don't see why you felt a need to start a thread without providing some real examples using your Hierarchical Bayes Model.

I don't try to predict speed figures but, instead eyeball the horse's previous ratings and attempt to determine whether there is a reason(s) for any improvement or regression in the prior races since, there are a number of reasons for changes. I'm one of those people who believes most numbers mean something.

How does your model account for these changes? And, again are you the grad student who knows nothing about horse racing?

[ReplayRandall]

Originally Posted by Greyfox

-

We're not stupid and know that Bayes is a mathematical approach to certain decision making situations.

Show us how you can get a better speed figure using it.

I'm sorry Greyfox, did I miscommunicate something in my post, just like Trifecta Mike?

[HUSKER55]

I tried to read that link and I decided to start drinking early.

Unless someone is going to put up some examples and explain things for us mere mortals I don't see this going anywhere for a lot of us.

[Greyfox]

Originally Posted by ReplayRandall

I'm sorry Greyfox, did I miscommunicate something in my post, just like Trifecta Mike?

Sorry. My mistake. It's Trifecta Mike who is trying to give us the lesson on Bayes.

[Sapio]

Originally Posted by Greyfox

Sorry. My mistake. It's Trifecta Mike who is trying to give us the lesson on Bayes.

Hi Greyfox,

I don't see where TM is trying to give a lesson on Bayes. The Bayes comment was a response to a poster's question, which really has nothing to do with this thread.

I agree with Robert Fisher's posts and Magister Ludi's post.

Thomas Sapio

[Greyfox]

Originally Posted by Sapio

I agree with Robert Fisher's posts and Magister Ludi's post.

Thomas Sapio

Both posters comments are highly debateable and in the absence of several concrete examples are not worth debating.
It is a simplistic overgeneralization, often faulty, to believe:
"Perhaps there is a cycle of equine performance consistency. W3 could be the minimum performance consistency point."