Live Handicappers' Exposition: Part 2

[cj]

Originally Posted by eurocapper

I suspect there is a tendency for authors to throw out what they have used before (when it's becoming unprofitable) while keeping the most recent angles for themselves. I think Beyer for example states he is nowadays playing to surface switchers. But how to bet is not a handicapping secret.

It is probably more likely they "throw out" what is working at the time they write, and after becoming popular it doesn't work as well.

[traynor]

Originally Posted by cj

Too deep for you?

You said the statisticians were experts, not me.

I think presenting "arguments" that require a lack of knowledge on the part of the audience to be considered valid are pointless. Rather than trying to convince an audience that lacks the knowledge to see the holes and deficiencies in your argument, you might do better to present that argument to those willing to take the time and exert the effort to point out the shortcomings. I am not one of those people. I don't especially care whether you believe small samples are predictive or not, or whether you believe that you can "make them work" because of some inside knowledge.

The basic fact is that the underlying premise is absurd. Not my idea. The idea of just about everyone with more than a superficial acquaintance with the topic. Hence--my suggestion that you put together an argument to convince a statistician that your premises are correct. Hopefully, one that does not require uncritical acceptance of personal experience or subjective belief to be meaningful.

[Tom]

The only statistic that maters is how much money you make.
The idea you can't make money on short term/small samples is ridiculous.
It is done every day.

[DeltaLover]

Originally Posted by Tom

The idea you can't make money on short term/small samples is ridiculous.

Hopefully this 'idea' can be analytically proven...

In other words it is a Popperian statement.

Can you please state the underlined theory and the related evidence ?

[cj]

Originally Posted by traynor

I think presenting "arguments" that require a lack of knowledge on the part of the audience to be considered valid are pointless. Rather than trying to convince an audience that lacks the knowledge to see the holes and deficiencies in your argument, you might do better to present that argument to those willing to take the time and exert the effort to point out the shortcomings. I am not one of those people. I don't especially care whether you believe small samples are predictive or not, or whether you believe that you can "make them work" because of some inside knowledge.

The basic fact is that the underlying premise is absurd. Not my idea. The idea of just about everyone with more than a superficial acquaintance with the topic. Hence--my suggestion that you put together an argument to convince a statistician that your premises are correct. Hopefully, one that does not require uncritical acceptance of personal experience or subjective belief to be meaningful.

How is it absurd? The proof is easy. Nearly every long term "statistic" or a combination of a few or even several that shows a profit in the past WILL NOT show a profit in the future. The win percentage may very well remain the same, but the prices will shrink. You can take that to the bank. That is real life in combination with what you call computer handicapping.

Aren't you the guy that said you could bet all your 3rd choices at 4 to 1 (or whatever odds) because you knew they won at a higher rate without breaking down the set to see, if in fact, those that went off at 4 to 1 were actually profitable alone?

[Tom]

Originally Posted by DeltaLover

Hopefully this 'idea' can be analytically proven...

In other words it is a Popperian statement.

Can you please state the underlined theory and the related evidence ?

It is proven by the increase in bankroll.
You play statistics. I'm play horses.
This is the real world, not a text book.

[DeltaLover]

Originally Posted by cj

How is it absurd? The proof is easy. Nearly every long term "statistic" or a combination of a few or even several that shows a profit in the past WILL NOT show a profit in the future.

This could be truth.

What you are missing here is that computer handicapping does not necessarily consists of an array of finite state machines.

In its more advanced levels computer handicapping evolves to an expert system with no reliance to concrete patterns that happened to be profitable in the past having instead the ability to take stochasticity in consideration resulting to a composing tier that makes decisions in a more intuitive fashion..

[cj]

Originally Posted by DeltaLover

Hopefully this 'idea' can be analytically proven...

In other words it is a Popperian statement.

Can you please state the underlined theory and the related evidence ?

It is a parimutuel game. You have to catch on before the public does. Good luck doing that with a 10 year database. Sure, you can find some things, particularly if you use your own, not publicly available, data. But the vast majority of things you find will not hold up going forward because the public already knows by the time you find them.

[cj]

Originally Posted by DeltaLover

This could be truth.

What you are missing here is that computer handicapping does not necessarily consists of an array of finite state machines.

In its more advanced levels computer handicapping evolves to an expert system with no reliance to concrete patterns that happened to be profitable in the past having instead the ability to take stochasticity in consideration resulting to a composing tier that makes decisions in a more intuitive fashion..

I am fully aware of what computers can do. I also know they can only do exactly what a human tells them to do, nothing more, nothing less. They can just do it faster and more accurately than a human can.

If a person doesn't program a computer to catch short term situations that are profitable before the public does, they are missing out on the majority of limited opportunities available in this game.

[DeltaLover]

Originally Posted by cj

It is a parimutuel game. You have to catch on before the public does. Good luck doing that with a 10 year database. Sure, you can find some things, particularly if you use your own, not publicly available, data. But the vast majority of things you find will not hold up going forward because the public already knows by the time you find them.

It is not possible to add a layer of indirection having this 10 year db instead of looking for actual profitable apporaches to detect the rate of their adoption from the public and the frequency of their generation?

[cj]

Originally Posted by DeltaLover

It is not possible to add a layer of indirection having this 10 year db instead of looking for actual profitable apporaches to detect the rate of their adoption from the public and the frequency of their generation?

Sure it is. But that isn't what I was arguing in the beginning. I was arguing with the person that said it is absurd to think small samples can be predictive. You seem to agree with me if I'm understanding your questions.

[Robert Fischer]

Originally Posted by traynor

I think presenting "arguments" that require a lack of knowledge on the part of the audience to be considered valid are pointless. Rather than trying to convince an audience that lacks the knowledge to see the holes and deficiencies in your argument, you might do better to present that argument to those willing to take the time and exert the effort to point out the shortcomings. I am not one of those people. I don't especially care whether you believe small samples are predictive or not, or whether you believe that you can "make them work" because of some inside knowledge.

The basic fact is that the underlying premise is absurd. Not my idea. The idea of just about everyone with more than a superficial acquaintance with the topic. Hence--my suggestion that you put together an argument to convince a statistician that your premises are correct. Hopefully, one that does not require uncritical acceptance of personal experience or subjective belief to be meaningful.

There's a difference between not being able to statistically prove that a small sample is significant, and claiming that a small sample can not be significant. This is not emphasized enough.

[DeltaLover]

Originally Posted by cj

Sure it is. But that isn't what I was arguing in the beginning. I was arguing with the person that said it is absurd to think small samples can be predictive. You seem to agree with me if I'm understanding your questions.

What constitutes a small or large sample is relative to the what we try to do.

In general a larger sample increases precision of estimates.

Assuming independent variables following the same distribution they will converge to the expected value as the size of population is growing.

The size of the sample is a function of the desired significance level.

For example for chi square we require minimum sample size for each category at least 5.

Here you can see on on line sample size calculator:

http://www.surveysystem.com/sscalc.htm

[cj]

Originally Posted by DeltaLover

What constitutes a small or large sample is relative to the what we try to do.

In general a larger sample increases precision of estimates.

Assuming independent variables following the same distribution they will converge to the expected value as the size of population is growing.

The size of the sample is a function of the desired significance level.

For example for chi square we require minimum sample size for each category at least 5.

Here you can see on on line sample size calculator:

http://www.surveysystem.com/sscalc.htm

For about the 100th time, I don't need a Stats tutorial. I've taken all the classes and done very well in them. The problem with stats is in their application. You either use them well, or you don't. Stats are misused in sports all the time, and racing is no different. This is especially true because of the parimutuel nature of the game.

I am not saying statistics can't be used to help win money. What I am saying is you better have a very deep knowledge of thoroughbred horse racing before you use them, and your knowledge of the game better evolve as the sport changes, because it always is.

[DeltaLover]

Originally Posted by cj

Stats are misused in sports all the time, and racing is no different. This is especially true because of the parimutuel nature of the game.

agree 100%

Originally Posted by cj

What I am saying is you better have a very deep knowledge of thoroughbred horse racing

Obvously you need some domain specific knowledge but its necesary depth of it is debatable and become the topic of a separate thread since we hijacked this one for good...