Putting an end to Beyer type speed ratings

[traynor]

Originally Posted by classhandicapper

I have 3 problems with modeling using stats or trying to improve on Beyers.

1. To mimic my thinking, I'd have to build a LOT of models.

For example, IMO:

The trainer factor is way more important in claiming races than in Graded Stakes.

Pedigree is more important in races with lightly races horses trying new things or making their first start than in races for older experienced horses.

Early speed tends to be more important at the lower end of the class structure.

etc...

That would be a LOT of models before I was done.

2. When it comes to ground loss, pace, bias and other trip questions, most of the time I can't quantify the impacts. Sometimes I believe the same exact trip impacted different horses by significantly different amounts. Even subjectively, I don't try to put an exact number on these things.

I just say to myself, horse X is better/worse than he looks on paper or horse X ran much better than horse Y with the same trip and stuff like that. It's hard to model "better" when "better" can be a couple of lengths or being eased depending on who won/lost a duel for example.

3. Occasionally, the major question in the race is something I don't have a stat on. So I have to intuit the answer based the specific case.

I would love to be able to push a button and have a computer spit out some odds lines etc.. But it just feels to me like computers are way better for doing research and for information retrieval than coping with the complexities of handicapping.

Appropriate data layering is an important (and often overlooked) aspect of data analysis.

[raybo]

Originally Posted by barn32

I've been out of town visiting with my mom. When my dad died in April she got herself a little dog for companionship. He follows her around everywhere she goes. The dog loves my sister. But the dog hates men. He barks incessantly at them and will not stop. He also attacks and bites at their heels.

I told my mother that by the time I leave (one week) I would have that dog eating out of my hand and sleeping in my lap.

Impossible she replied.

It only took two days.

Do you see how this story relates to Trifecta Mike?

Sorry, I must be dense, I don't get it.

[Greyfox]

Originally Posted by raybo

Sorry, I must be dense, I don't get it.

The story is conveying that someone tells you that he can do something and then he won't tell you how he does it.

[raybo]

Originally Posted by Greyfox

The story is conveying that someone tells you that he can do something and then he won't tell you how he does it.

Well, TM is only one of many here that fit that description, and if their reason for not doing so is to protect their edge (if they really have one), then I can't argue with them. I certainly would never offer my edge to anyone, but I will tell them the mechanics of doing something similar. That would at least give them a starting point from which to do their own work, without damaging my own edge. Invariably, IMO, the person who creates something complex will be able to leverage it in ways that other users cannot, because they didn't design and implement it, knowing everything that is involved. So, even if you tell someone basically how to do something, they still have to do it, and will not do exactly what you have done.

[Cratos]

Originally Posted by classhandicapper

I have 3 problems with modeling using stats or trying to improve on Beyers.

1. To mimic my thinking, I'd have to build a LOT of models.

For example, IMO:

The trainer factor is way more important in claiming races than in Graded Stakes.

Pedigree is more important in races with lightly races horses trying new things or making their first start than in races for older experienced horses.

Early speed tends to be more important at the lower end of the class structure.

etc...

That would be a LOT of models before I was done.

2. When it comes to ground loss, pace, bias and other trip questions, most of the time I can't quantify the impacts. Sometimes I believe the same exact trip impacted different horses by significantly different amounts. Even subjectively, I don't try to put an exact number on these things.

I just say to myself, horse X is better/worse than he looks on paper or horse X ran much better than horse Y with the same trip and stuff like that. It's hard to model "better" when "better" can be a couple of lengths or being eased depending on who won/lost a duel for example.

3. Occasionally, the major question in the race is something I don't have a stat on. So I have to intuit the answer based the specific case.

I would love to be able to push a button and have a computer spit out some odds lines etc.. But it just feels to me like computers are way better for doing research and for information retrieval than coping with the complexities of handicapping.

The computer calculates and the human mind imagines; together they make a wonderful team.

[PaceAdvantage]

Originally Posted by barn32

I've been out of town visiting with my mom. When my dad died in April she got herself a little dog for companionship. He follows her around everywhere she goes. The dog loves my sister. But the dog hates men. He barks incessantly at them and will not stop. He also attacks and bites at their heels.

I told my mother that by the time I leave (one week) I would have that dog eating out of my hand and sleeping in my lap.

Impossible she replied.

It only took two days.

Do you see how this story relates to Trifecta Mike?

Yeah, but the dog was worth the effort. The jury is still kind of out on our pal TM.

[Sapio]

Originally Posted by PaceAdvantage

Yeah, but the dog was worth the effort. The jury is still kind of out on our pal TM.

Hi PaceAdvantage,

I know I am dense in these matters. Can you explain this to me?

Thomas Sapio

[HUSKER55]

That is a polite way of saying small fry's like us should stay out of this conversation.

[MJC922]

Trifecta Mike has shared some unique insights. He certainly got me thinking out of the box with regard to modeling and I'm grateful for that.