Out of sample testing of patterns

I didn't invent anything, just saw your poor garbage and replied. What you did is wrong, plain and simple. Because you still fail to understand, I'll even help you further. Blue dots are days.

attachment.php


I made it clear in my earlier post that you either lied or you are a fool. So technically, I agree with you, I don't know if you lied. You might simply be an idiot. I'm sorry if the latter is the case.
 

Attachments

Quote from Code7:

I made it clear in my earlier post that you either lied or you are a fool. So technically, I agree with you, I don't know if you lied. You might simply be an idiot. I'm sorry if the latter is the case.
More...

No, you called me a liar several times, along with other names, and my lawyer is on this matter now.

Now, you always refer to abstract things like "large collections of random trades" and your latest manufactured example to try to confuse me. Mind you, I have been testing and trading systems for over 20 years. Referring to the Harris' patterns, you still insist despite the fact that you do not even have the software to check the patterns out. I generated code for each of the patterns found by APS in the in_sample period and backtested all patterns in the out_of_sample. The aggregate results are:

Sum winners = 235.65 points
Sum losers = 187.48 points
Net profit = 48.17 points

Profit factor = 1.26

Now, you insist that some "large average of random trade collections" would either match or outperform this system of 8 patterns in the same period, specifically I also used the same period as Harris.

You have not presented any such system. You just talked vaguely, like someone who has never traded a single contract but still shows his ego in trading forums. Traders do not trade some "large average of random trade collections" because commissions kill them. By comparing to that you proved you have no actual experience with trading.

If you do not post a system that satisfies your claims based on random trades that has the same or fewer number of trades as Harris's system I will seriously consider going ahead with the lawsuit for all the insults you have thrown. I have also sent an email to Harris about you and your constant attempts to defame his software.
 
I find it interesting how defensive you're getting over this intradaybill, and, I'm having a hard time taking your threats about legal action very seriously given how legally, being called a liar on a internet board means squat and code7 isn't doing anything other than poking holes in certain claims...

That said, I'm sure you could find a slimy "lawyer" to deal with this issue, but, the only reason that would be necessary was if you had a vested interest in pushing this product, i.e. you are in a position to profit. So, based on your level of defensiveness, are we to conclude you actually do have a vested interest in this product?

My advice would be to combat this with some sound math. So far, I see nothing on your end that eliminates the correlation = causation fallacy. In fact, I think your understanding of random entry is slightly off, please refer to this thread (and read it all the way through despite the BS):

http://www.elitetrader.com/vb/showthread.php?s=&threadid=179402
 
Quote from Mike805:

That said, I'm sure you could find a slimy "lawyer" to deal with this issue, but, the only reason that would be necessary was if you had a vested interest in pushing this product, i.e. you are in a position to profit. So, based on your level of defensiveness, are we to conclude you actually do have a vested interest in this product?
More...

…and may I ask what is your vested interest in defending Code7? Are you a collaborator of Code7. Read my posts carefully. I am arguing about system design. Specifically I am arguing against the method Code7 used to measure the effectiveness of a system. I am arguing that (A) "large collections of random trades" and (B) a specific trading system with specific entries/exits cannot be compared because A has hidden commission cost and more importantly it does not represent a specific system of any kind.

Code7 argued against that accusing me of being a liar. So I conclude Code7, his collaborator Lizardo and you are intentionally making stupid arguments. Why? I do not understand why. Maybe you have a vested interest in attacking the particular software I took the examples from. Someone already provided some information about Code7 I am checking out.
My argument is purely related to trading systems. Forget about the software used or anything else. Nobody will come here to push any software because it does more harm than good. The heat is instant and you know that.
This is the argument:
You have a system with an actual PF = 1.30 and someone comes and says after the fact that that was not a good system because a "large collection of random trades" could have achieved the same or even better result. This is basically the argument Code7 is making.

All I am arguing is that is not a sound argument and I used a specific example for that I can work on and analyze.

What is your vested interest?
 
Quote from intradaybill:

I just found you vested interest

http://www.elitetrader.com/vb/showthread.php?s=&postid=2824459#post2824459

Do you hold TS stock? What is your vested interest in that product?
More...

Bill, I have no vested interest in TS other than the fact that I use the product and believe there is value in the data package. A data source is a very different animal from a system design tool that performs a type of data mining.

I have never collaborated with code7 and do not have any reason to attack you or your choice of system design tools. However, I am not convinced based on your replies that you are understanding the root issue that code7 brought up in his very first point, i.e. the random entry condition and its appearance of adding value to a system.

Its very easy to add a set of random conditions to a system and get a much much better hindsight equity curve. I can provide a dozen or so systems that do this and they look tradeable and will pass forward sample tests. In fact, if you bothered to read the thread I provided in my prior post, you'll see one of those systems.

There is a major flaw here and it has a lot to do with the designer's choices: how do you know the condition you've provided adds value randomly (i.e. the correlation = causation fallacy) versus discovers a fundamental market behavior ?

Mike
 
It appears there is much communicating at cross purposes here.

Intra - "large collections of random trades" doesn't mean "lots of trades", it means "trades from lots of random systems". Specifically, in this case, it means randomizing the long-only entries with similar hold & exit conditions to the APS trades. And Code is right - the APS signals will not outperform such a random selection. This is very easy to test.

Put very simply - the value in the APS "patterns" is that they are long-only during a long bull run, and have nothing to do with the patterns themselves.

If you're going to spend much time at this, what you should be focusing on is how to determine when you should be running long-only or short-only trades. :) Because without that, you have nothing.

Cheers.
 
Quote from Random.Capital:

It appears there is much communicating at cross purposes here.

Intra - "large collections of random trades" doesn't mean "lots of trades", it means "trades from lots of random systems". Specifically, in this case, it means randomizing the long-only entries with similar hold & exit conditions to the APS trades. And Code is right - the APS signals will not outperform such a random selection. This is very easy to test.

Put very simply - the value in the APS "patterns" is that they are long-only during a long bull run, and have nothing to do with the patterns themselves.

If you're going to spend much time at this, what you should be focusing on is how to determine when you should be running long-only or short-only trades. :) Because without that, you have nothing.

Cheers.
More...

Thanks, I understand but it is equivalent to a sense, trades from a large number of systems = large number of trades. More importantly, it says nothing as far as actually trading. It is a theoretical concept with no value to trading. You can only determine if such system did better only in hindsight. On the other hand you need a system to trade. What would you do if you wanted to trade? Would you take the Harris system with a PF ~ 1.3 or wait after 6 years to claim that the Harris system did as well as some large selection of random trades from a large collection of random systems (something that remains a claim and not proven by the way). Or worse, trade randomly?

I hope you get my point. I think his argument (Code7) is a strawman. Otherwise we should all stop trading because after 6 years someone can say "hey you stupid guys, you should have thrown darts instead of using systems".

You know, there are many problems with probability theory, especially the axiomatic versus the frequency defintions. What he claims is that a large number of trades will converge to the average expectancy but the issue is, how long that sample should be to be confident it will? Theoretically, you can take 500 trades form a large collection of random systems and still not hit the average expectancy. It is all a theoretical number like the probability of heads of a fair coin. You could toss 1000 tails before head comes up.
 
Quote from Mike805:

Bill, I have no vested interest in TS other than the fact that I use the product and believe there is value in the data package. A data source is a very different animal from a system design tool that performs a type of data mining.
Mike
More...

TS is a data source? For God's sake. TS is a trading system development and execution tool. You can do data mining with TS. Every time you develop a system you are essentially doing some sort of data mining.

Intradaybill on the other thread did not praise a tool against another as you did. What makes you think that you have the right to praise your own tools, especially against other tools, but when other posters talk about their tools you can freely accuse them of having a vested interest in them?

I think your behavior is quite peculiar in this respect.
 
Quote from alexandermerwe:

TS is a data source? For God's sake. TS is a trading system development and execution tool. You can do data mining with TS. Every time you develop a system you are essentially doing some sort of data mining.

Intradaybill on the other thread did not praise a tool against another as you did. What makes you think that you have the right to praise your own tools, especially against other tools, but when other posters talk about their tools you can freely accuse them of having a vested interest in them?

I think your behavior is quite peculiar in this respect.
More...

Well, I use TS primarily as a data source and Amibroker as a system design tool, of which I also think very highly.

What strikes me is how defensive bill is over this issue and simply put, I want to see the math. Personally I don't believe much in using patterns, but, if one has sucess with it then by all means, carry on.

When presented with some difficult questions and tough criticisms, bill seems to get agitated very easily and threatens legal action.... and while that's not "praising" as you refer to it, then what exactly is it?

Mike
 
You must log in or register to reply here.
Back
Top