Question on using compiled factors...

[Mulerider]

Originally Posted by mikesal57

So you now have the dilemma here that I posted....

You can leave BP in

or

Leave BP out....

I ran into your dilemma when I tried to formulate my own "class" ranking...just substitute Bris's "Current Class" rankings for Prime Power. Without knowing Bris's metrics and weightings for it, there was the danger of redundancy if I used it in conjunction with my own factors.

[headhawg]

This is an interesting thread. I am working on a new program and this is one of the things I am trying to improve upon. I am not a stats or math wiz but it seems to me that factors that are intercorrelated are not ideal. Perhaps developing a power factor that performs as well or better than Prime Power would be good as you would know what combination of factors went into it. You could then use other factors that have less or no overlap with what is in your own power number.

Of course, if you are looking for higher priced horses you need to control the use of factors that correlate too highly with the tote board. I don't want to sound like a shill, but Dave Schwartz's stuff has really given me a lot to think about. I am using some of his ideas along with some of my own modifications in the new program.

[mikesal57]

Originally Posted by headhawg

This is an interesting thread. I am working on a new program and this is one of the things I am trying to improve upon. I am not a stats or math wiz but it seems to me that factors that are intercorrelated are not ideal. Perhaps developing a power factor that performs as well or better than Prime Power would be good as you would know what combination of factors went into it. You could then use other factors that have less or no overlap with what is in your own power number.

Of course, if you are looking for higher priced horses you need to control the use of factors that correlate too highly with the tote board. I don't want to sound like a shill, but Dave Schwartz's stuff has really given me a lot to think about. I am using some of his ideas along with some of my own modifications in the new program.

this is what people dont realize.....they can create 1000 factors but once they start using something like Bris Prime....they are doomed to get a ROI in the red and all they created went to waste

This is why I ask you HH , if your program can provide a CSV output file ...so that numbers/factors can be validated

[headhawg]

My intention is to provide csv output, yes. I am still working on factor weights, and also feature design -- what kind and how many options for the user.

[Dave Schwartz]

Headhawg,

Of course, if you are looking for higher priced horses you need to control the use of factors that correlate too highly with the tote board. I don't want to sound like a shill, but Dave Schwartz's stuff has really given me a lot to think about. I am using some of his ideas along with some of my own modifications in the new program.

Thank you. That is my primary goal.

The thing about these compiled factors, as Mike called them... Sure, they correlate strongly with the tote board... BUT...

Collectively, so does everything else!

We all know that there are no single factors that stand out on their own as being kick-ass.

So, along comes a genius who says, "I have a completely new 'object' that points to who is likely to go to the front. My prediction is right 57% of the time. Since I am the only one who has it, I should do very well."

Au contraire. Sure, his approach for picking the likely 1st call leader may be very unique, but what he forgets is that a guy using Quirin Early Speed Points will get that same horse 48% of the time. (I made these numbers up.)

Since the world is very aware of the Q-ES-Pts, the only thing he's done is created a complex version of those points which, according to my example, would be 20% better.

His object would still correlate!

BRIS Prime Power is the same way: It is not actually a single factor. It is made up of many factors. If you use any of those factors that make up BPP...

OR

... you use factors that correlate with the factors that BPP uses, then YOUR OBJECT CORRELATES.

In that case, you've gained nothing.

BTW, in the above example of "20% better," the question is, "Will 20% better than Q-ES-Pts be enough to make a significant difference?"

Probably not.

It's like trying to build a better speed rating. Can it really be SO MUCH BETTER that it takes you beyond break even?

When you take into consideration that everybody uses speed ratings, it makes it even harder.

Remember that when I say, "everybody" I mean even the people who say proudly, "I use no speed ratings! I use class."

Guess what? "Class" (however you define it) WILL correlate with speed ratings. Therefore, it will correlate with Prime Power.

See why the game has become so difficult?

That's why I've personally become so very contrarian: It is the only chance.

Alas, it is a lot of work to be contrarian.

But soon it won't be.

Just my opinion.
Dave

[mikesal57]

Originally Posted by headhawg

My intention is to provide csv output, yes. I am still working on factor weights, and also feature design -- what kind and how many options for the user.

HH

Flow is....


Developed Factor=======> Data/Results ========> Weights


Dave....

You cant live without Bris Prime......and tossing it is no better

So

What the alternative????

To find "Why/How Bris Prime Fails"

You cant beat it in every race but can you pick your spots?

Mike

[Dave Schwartz]

Originally Posted by mikesal57

HH
Dave....

You cant live without Bris Prime......and tossing it is no better

So
What the alternative????

To find "Why/How Bris Prime Fails"

You cant beat it in every race but can you pick your spots?

Mike

I contend that the only real value BPP and PSR have is to predict what the odds SHOULD BE.

That is what I am betting into!

The alternative is to use an approach to handicapping that the whales cannot and will not use, and then figure out do you bet the horses that come up as plays or do you bet against them.

It is a binary choice.

This is what the HSH Handicapping Tiers seminar was all about.

[Tom]

Just a thought on BPP - it is based on recent races at similar distances and surfaces, so a low PP rating on a horse who has been running race not similar to today, ie, a freshened horse wit two sprint preps dropping today and stretching, the PP may be too low to the horse's real fit today. Or a Linda Rice turf horse who ran on dirt all winter and now gets back on the lawn.

If you have a database of BRIS, you could look up his PP for a race in the past where the horse was racing close to todays set up.

[46zilzal]

There are FEW universal factors....Aqueduct may have specifics that Thistledown does not. Bull rings has factors that 9 furlong ovals do not etc etc

Turf dirt, sprint routes....AND it is important to review periodically to see just how well they work.

An amazing factor is two best lines ever.

[mikesal57]

Originally Posted by 46zilzal

There are FEW universal factors....Aqueduct may have specifics that Thistledown does not. Bull rings has factors that 9 furlong ovals do not etc etc

Turf dirt, sprint routes....AND it is important to review periodically to see just how well they work.

An amazing factor is two best lines ever.

Just to clarify a little Tim....2 best lines in the PP's

There was an approach that The Hat (Bradshaw) had which was to match-up horses BEST line with each other.....It didn't get a lot of winners but when it did hit ...prices were nice..

and YES....tracks need to be handicapped individually

[46zilzal]

[QUOTE=mikesal57;2468431]Just to clarify a little Tim....2 best lines in the PP's

There was an approach that The Hat (Bradshaw) had which was to match-up horses BEST line with each other.....It didn't get a lot of winners but when it did hit ...prices were nice..

/QUOTE]

NOT quite the same...

[headhawg]

Originally Posted by Dave Schwartz

...BTW, in the above example of "20% better," the question is, "Will 20% better than Q-ES-Pts be enough to make a significant difference?"

Probably not.

It's like trying to build a better speed rating. Can it really be SO MUCH BETTER that it takes you beyond break even?...

But what if one could find multiple "unique" objects that were 20% better? Is that not possible given that the data has been sliced-and-diced so many times that there are no more of them?