Genetic Programming

[alcodemar]

Good day,

I am new to this forum and have found numerous threads that are very interesting. I would be interested to hear from members about their experience with Genetic Programming (GP) software on the market. I am interested in building trading systems using TS but I have little skills in programming. Software such as Adaptrade and Trading System Lab (TSL) claim they can achieve it building robust systems. Any comments, suggestions?

Thanks,

 

[avano]

I tried adaptrade and couldn't get anything that makes sense. Most strategies it got were like chandelier exit systems (enter randomly at the open and exit with some complex function of ATR or some triangular ma). Also, every time I ran the same settings I got different results which shows randomness. There is another program I am considering called Price Action Lab which generated EL code but each time I ran the demo with same settings I got the same results.

 

[alphadude]

Adaptrade; like any GP tool; is not bad. It does have the basic blocks of GP. There is another tool called Jgap; but Adaptrade is perhaps the leader in its domain.

Having said that; my advice is stay away from machine learning stuff. The search space is so big; and the world is so complex; that statistically it is impossible to achieve any success by taking shortcuts.

find yourself a mentor; and roll up your sleeves.

looking for the nazi secret box wont get you anywhere. Hungarians had it; and only made 89 copies; and all got sold. if you find one for sale; buy it.

 

[avano]

Adaptrade; like any GP tool; is not bad. It does have the basic blocks of GP.

Have you used it or you just say so because you have to build up the post count?

Do you have results to discuss for E-mini for example?

 

[alphadude]

Have you used it or you just say so because you have to build up the post count?

Do you have results to discuss for E-mini for example?

unlike you i don't claim to be foremost expert in system dev and don't have any nazi secret box for sale.

as to Adaptrade; indeed i did try the software for a couple of weeks; and I was very impressed with it.

why? i did dwell into GP a couple of years ago; and built my own app to do the optimization. When I come across Adaptrade; i threw away my stuff; as this soft was superior.

It is also flexible in the following ways:
- has many of the domain indicators, price patterns..etc
- allows you to use your own custom indicators
- spits out TS easy language; which is understood by human reader
- has in-sample/out of sample tests
- etc.

i personally have tried many AI stuff; Fuzzy Logic, NN, GP; and given up on them; as I found simplicity and domain knowledge to be the key in trading success. There are other nitty gritty details beside having a working strategy; that are important in this process.

Even if a GP app spat out a formula that did well in testing and forward testing; if I don't make sense of its logic and causality; i won't bet any money in it.

That is the problem with many machine learning stuff; you don't understand the logic behind the system; and hence you won't be able to know when the system begin to die.

GP is an exception; in that the generated formula can be read. The issue with it though; is you will find yourself spending too much time in deciphering garbage formulas spat out by the optimization program. The efforts and time spent is not worth it. It is akin to looking for gold nuggets in a river stream. I would rather be learning about trading domain and building up from there.

Only in some scenarios GP could be useful to you; and that is when you are already trading successfully; and you absolutely know what you are doing. You can then use GP to search a narrow space for some optimization problems.

so will I use Adaptrade in the future; possibly yes. It is added to my wishlist; which contains many many other *sexy things to try on free time*. am simply too busy with trading my system that is based on solid foundation.

so my advice again is; unless you are already trading successfully; stay away from machine learning stuff. instead; build up your inventory of knowledge; block by block; and most of all; build up you thousands of screen time. no substitute for that.

Human brain is much more powerful than GP apps. If you feed it the right input data; it will output the right data.

 

[tovim]

Try this one Eureqa | Cornell Creative Machines Lab at least it is free and it was fun for me to play with it .

 

[avano]

as to Adaptrade; indeed i did try the software for a couple of weeks; and I was very impressed with it.

You are either a college dropout who gets impressed with arcade programs or a snake oil salesman. I tried that program and so a few others I know and the results were for laughs. Unlike you and your general talk, this is a specific example of a long entry condition that I got from Adaptrade:

L[1] <=c[14]

Now, if you do not understand that this is random, then it is because you know nothing about this business. You sound like a salesman to me. I have to put you on ignore. Please do not answer any of my posts. You have nothing to contribute. Bye, did you listen to me? Bye.

 

[alphadude]

You are either a college dropout who gets impressed with arcade programs or a snake oil salesman. I tried that program and so a few others I know and the results were for laughs. Unlike you and your general talk, this is a specific example of a long entry condition that I got from Adaptrade:

L[1] <=c[14]

Now, if you do not understand that this is random, then it is because you know nothing about this business. You sound like a salesman to me. I have to put you on ignore. Please do not answer any of my posts. You have nothing to contribute. Bye, did you listen to me? Bye.

son; obviously you haven't read my review of the software above; and pros/cons of using GP; or you don't know sh*t about GP to start with. your reasoning above is so stupid; that you actually don't know how to handle the output of the app.

let me proove you are naive:

this formula; L[1] <=c[14]; is a a common mean reversion strategy; that opens a long position if yesterday low price; is lower than the close 14 days ago. It is not a random entry. Variants of this system is well known to the quant community for years.

the GP optimization algorithm is not deterministic. meaning; each time you run the optimization; you get different formula. This is by design. the GP algo has a random component to it to do the genetic mutation.

I suggest you start with a book on GP before you open your mouth again on this subject.

am still waiting for your future post to sell your nazi software. tick tick tick...

 

[avano]

son; obviously you haven't read my review of the software above; and pros/cons of using GP; or you don't know sh*t about GP to start with. your reasoning above is so stupid; that you actually don't know how to handle the output of the app.

I suggest you start with a book on GP before you open your mouth again on this subject.

am still waiting for your future post to sell your nazi software. tick tick tick...

GP is a way to curve fit you idiot retard. Not only you are a retard, you look like one in the picture you chose. No matter how you insult others, everyone understands you are a snake oil salesman when you say you were impressed with adaptrade. Here what some its users say you moron:

Strategies quickly fade: Strategies quickly fade - Adaptrade Builder | Google Groups

Is Builder simply fitting the trend? Is Builder simply fitting the trend? - Adaptrade Builder | Google Groups

 

[DionysusToast]

GP is a way to curve fit you idiot retard. Not only you are a retard, you look like one in the picture you chose. No matter how you insult others, everyone understands you are a snake oil salesman when you say you were impressed with adaptrade. Here what some its users say you moron:

Strategies quickly fade: Strategies quickly fade - Adaptrade Builder | Google Groups

Is Builder simply fitting the trend? Is Builder simply fitting the trend? - Adaptrade Builder | Google Groups

Errr....

I'm confused.

You mention curve fitting, which virtually guarantees you do not have something that will work moving forward...

and you post some reviews by people who are saying it doesn't do a good job?

Why would you be expecting anything other than poor results moving forwards?

 

[alphadude]

GP is a way to curve fit you idiot retard. Not only you are a retard, you look like one in the picture you chose. No matter how you insult others, everyone understands you are a snake oil salesman when you say you were impressed with adaptrade. Here what some its users say you moron:

Strategies quickly fade: Strategies quickly fade - Adaptrade Builder | Google Groups

Is Builder simply fitting the trend? Is Builder simply fitting the trend? - Adaptrade Builder | Google Groups

interesting that you keep projecting yourself; and your failures.

 

[alphadude]

Errr....

I'm confused.

You mention curve fitting, which virtually guarantees you do not have something that will work moving forward...

and you post some reviews by people who are saying it doesn't do a good job?

Why would you be expecting anything other than poor results moving forwards?

the gullible doesn't know what he is talking about.

obviously he is looking for the secret nazi formula.

 

[avano]

the gullible doesn't know what he is talking about.

obviously he is looking for the secret nazi formula.

Nutcase alphadude. Sleazy snake oil salesman.