My goal is to write a expert without any parameters... so basically i have one indicator and all the other parameters are depending on the output of that one. Trough this system i get a full dynamic system which is (in theory) able to "self-adjust" to all markets. Of course this is not smart but striclty rules.
My next EA i think will be a NeuralNet EA, but i don't want to go the direct way (getting signals out of the NN). My idea is to build a EA trough more NN, i.e. one NN to recogice different markets, one to configure a few indicators an finally 3 or 5 NN wired trough some fuzzy logic to pull out signals.
But i have not begun coding. so i can say nothing about the performance.
to your post:
a) Well, trading the past is easy, normaly not profitable in the future..
b) Depending on the type of forecast maybe profitable
c) Since it's statitical it's also some kind of optimization.
d) Human Brain is still the most powerfull. go for brain!
//z
The challenge, in my opinion, is that all four options could produce lackluster results and you won't know if it is an instrinsic failure of the method/strategy itself or merely limitations introduced by way of its specific implementation.
Likewise on the flipside, it is all too easy and tempting to view short-term favorable results as indication of proof that a given strategy (and its specific implementation) is the way to go.
Achilles heel or Holy Grail.
Moby-Dick or Alchemy.
Nobody wants to realize in hindsight they are Captain Ahab or جابر بن حيان
@zzuegg: the implementation of Neural Networks would seem to me the most attractive. Do you really believe this technology is suited for trading where the rules of the game are infinite?
@ Phillip: I've come to realize that learning to Trade and Program is a never-ending process. In the End, I'd want a Smart EA to trade Like-Me as much as possible. What's the best path in your opinion for creating such system? As I could be doing this 20 years from now, I'm hoping to generate some outlook on smart EA's.
@zzuegg: the implementation of Neural Networks would seem to me the most attractive. Do you really believe this technology is suited for trading where the rules of the game are infinite?
Well, not that easy to answer. There are very different types of NN's. I think they work well as when you use them as a pattern indication indicator. I belive in NN's and so i decidet to program one...
@ Phillip: I've come to realize that learning to Trade and Program is a never-ending process. In the End, I'd want a Smart EA to trade Like-Me as much as possible. What's the best path in your opinion for creating such system? As I could be doing this 20 years from now, I'm hoping to generate some outlook on smart EA's.
Neural-net that uses your actual manual trading history to optimize and learn from. So go do manual trading for a good year, keeping good records of your trading history and then train a neural net on it.
So I've been thinking, lately with a few posts related to Smarter robots, what's Best approach for creating a smart Robot.. Please respond with your Option and Opinion. In my past surveys I didn't allow counter points. This time I would like counter opinions. Hopefully a little constructive criticisms wouldn't ruin the post.
Here are some options for taught.
a) Self_Optimizing EA's based on MT4 Optimization facilities.
b) Neural Network Optimizations based on traditional neural definitions.
c) Chaos/Fractal Geometrical patten recognition (including fabo and gann)
d) This is all BS, good ole human brain for me thank-you :)
Well if there's some other advanced technology you'll like to see on the list, please let me know in time so I can edit the options. Otherwise you can just list it as part of your comment.
NN / AI is certanly the direction that we should be headed.
With that stated, That's like a GPU on an 8086 processor.
Meaning that first we have to have better development tools in-order to create our code.
MTn ( where int n >= 4) is a discusting place to develop decent code. (1)
I suspect that many/most of us use something like Visual Studio to write/test/debug this, however I see, too, that many questions arise because simple debugging is not able to be done.
NN- Yes. Combined with Geo-P-R, Yes.
Self Optimizing - Only after HAL-9000.
First, we have nothing to 'model' self optimizing on.
Second, From the simplist game to the most complex missile systems, most applications have/allow the user to interact with it;
- Even if it is only to select the defaults.
- - - -
(1)
In days past we had PC-dos, IBM-DOS, Berkeley-DOS, R-DOS and and about 1.2 unique versions of a "DOS" for every university in the world.
Just Thank ( your creater here) for the ARPA net, lest we would never have communicated at all.
But DOS gave is the -Green on Black- Bulliten-Board System (BBS) such as Genisys, or Colossus, (or your own bbs here)
My point is that everyone used BBS's and thought they were the wave of the future. Today we have the same issue with MT4/5.
We need a better code development platform in-order to create powerful EA's - no matter what methodology they contain.
I hope this touches your points-of-topic.
@MXVC.VII.LCXVI
Maybe some revolutionary person would come along and develop an open source TMS (Trade Management System) kinda like what happened to Linux. MT5 includes OOP Object Oriented Programming and Multi-Treading. Does this Not put it on par with allot of todays Modern Programming Languages? In other words, what does Visual Studios have that it lacks? Better debugging tools for example?
Getting back to NN, it could be that allot of developers are not showing interest in this area because it's still just theoretical. Seemly NN would work well within a game of Chess where the moves and the confines of the chess-board are well defined. In trading however, the answer to buy---sell or sit would seem more like asking, does she love me or not. Like someone here pointed out, in the end you don't want to feel like you're running a fools errand. Whatever happened to all the buzz about chaos/fractal theories.
My approach would be to learn to trade first. Like learning to play chess, then developing a NN which follows the moves I've learned. To summarize my approach, I don't think the programmer who created Deep-Blue could beat a reigning Grand Master, but his NN could surly do the trick.
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
So I've been thinking, lately with a few posts related to Smarter robots, what's Best approach for creating a smart Robot.. Please respond with your Option and Opinion. In my past surveys I didn't allow counter points. This time I would like counter opinions. Hopefully a little constructive criticisms wouldn't ruin the post.
Here are some options for taught.
a) Self_Optimizing EA's based on MT4 Optimization facilities.
b) Neural Network Optimizations based on traditional neural definitions.
c) Chaos/Fractal Geometrical patten recognition (including fabo and gann)
d) This is all BS, good ole human brain for me thank-you :)
Well if there's some other advanced technology you'll like to see on the list, please let me know in time so I can edit the options. Otherwise you can just list it as part of your comment.