Jack H - To me, you are another loser like small bicycle.
Don't you ever ask yourself why you can't trade beside just talk and talk ?
Use computerized evolution to create trading system?
- Thread starter lighnintrade
- Start date
Jack H - To me, you are another loser like small bicycle.
Don't you ever ask yourself why you can't trade beside just talk and talk ?
You are basically describing the generative method of building trading systems. It is a common method in use. The other method has been termed scientific/deductive.
There are many programs that could be used for this purpose StrataSearch, PriceActionLab, Biocomp's products, possibly R, RapidMiner, Grail's System Builder, and ChaosHunter.
I imagine it would take months of work to get something worthwhile though and some degree of skill. Most of these methods are better for refining/tweaking sound trading strategies rather then creating something from scratch. Generative methods are very notorious for generating systems that produce unpredictable out-of-sample results.
There are many programs that could be used for this purpose StrataSearch, PriceActionLab, Biocomp's products, possibly R, RapidMiner, Grail's System Builder, and ChaosHunter.
I imagine it would take months of work to get something worthwhile though and some degree of skill. Most of these methods are better for refining/tweaking sound trading strategies rather then creating something from scratch. Generative methods are very notorious for generating systems that produce unpredictable out-of-sample results.
I think you are mixing a few things here, maybe because you haven't tried any of these.
StrataSearch and Price Action Lab create systems that traders are more familiar with when looking at them. StrataSearch is more of a permutation tool in my experience. It can create some strange rules. On the other hand, Price Action Lab's domain is restricted to price patterns to limit data mining bias as they claim. This one creates code for TS, WL4, and formula code for a few others.
ChaosHunter is more of a general statistical tool for forecasting. RapidMiner is a general data mining program that most traders may find hard to use so that they get something familiar to them out of it unless they have a degree in statistics.
Biocomp's product is AFAIK a black-box. I am not sure it is not optimizing as they claim.
The Grail strategy builder I think was bought by TS but IMO it is a walk-forward optimizer.
If you have the time and knowledge it worths trying to create yours. Otherwise, you may get one of these depending on your experience and knowledge of programming and statistics.
Tres bien...
No argument here.
Assuming one believes all systems are curve fits, part of the analysis can be "destructive testing" - screwing around with the parameters and seeing if it blows up or just detunes. We do this in engineering all the time (tolerance and failure analysis).
Run it until it breaks? Not a bad idea if it breaks gently and allows you to turn it off without getting sucked into a liquidity trap. LTCM comes to mind.....
