Overly Complex Coding and Trading Systems. - Page 3
Page 3 of 614 FirstFirst 1234 LastLast
Results 21 to 30 of 32

Thread: Overly Complex Coding and Trading Systems.

  1. #21
    Jure,
    It sounds like you have had some experience with NN's. I have not so if I say something that makes no sense bothers me, however, had a few questions from everything you stated. I am only asking about some NN that could have backward capabilities.


    You can not just make a lot of interconnected neurons, nourish it using raw cost (probably a window of previous prices) and anticipate anything useful to come out. Feeding it with multiple inputs (such as many indiors( probably derived from cost) will probably make things worse.

    --Now it sounds like the curse of dimensionality is something I must be afraid of, but signs do seem redundant because there's no executable actions. Would actual trading systems with functions that make executable actions out of information help? Kind of..clear upward and refine this information?


    There's also always the threat for the internet to overfit the data. The internet would literally incorporate the training set, while totally neglecting at predictions on new information.

    --Would you be able to incorporate the information but dump least powerful based on a defined set of parameters, while continuously accepting new data?


    I could go on, but ... the point is, there is a good deal of voodoo necessary to make anything related with NNs. A good deal of people believe a NN is some black box that magically learns to map background to future

    -- noted But watching as we know the long run is and always will be that. . Based from what claude appeared to be going to get was associated with probablility. I believe he had been saying before. . Speaking about the likelihood of the subsequent 3 bars assembly purchase or sell conditions. That would not be out of the domain of an NN would it?

    SO... sorry to clog the thread, you won't see me do it very often that's for certain. But claude says, a really intelligent man, constantly has more questions than answers. I am really eager to see where this goes and will tell I am gonna nead a life-jacket. Claude, Jure, Be-mac, it is going to be fun watching experienced minds woking together on this stuff.

    Fantasies of appreciation and Best of luck,
    Jax

  2. #22
    Very interesting work with your NNs Claude.

    I have been playing with some ideas that may use genetic calculations myself. Basically I got the idea from reading articles about Financial Labs that uses GA to achieve some fantastic results (24% annum/4 percent maxDD). My research currently pertains to basically creating various systems and then letting them loose where the GA selects for the best performing placing most weight on this, breeding successful systems and adapting to new market structure (mutations). You may want to check out literature about that.

    No breakthough ideas here (yet!) But its fairly intriguing.

  3. #23
    I understood that my previous post might have turned out a bit too negative. I wanted to address the holy grail thing, to clarify. NNs are the holy grail of automated trading just like Foreign Exchange is the holy grail of money making. To put it differently, the potential is excellent, but so is the effort necessary to actually make use of some of the potential.

    Claude:
    I wrote my own appliions. Nothing fancy, only a library of everything I accumulated through experimenting (I experimented with GA, yes). I did check out some software packages on the market, but I constantly found some limitiations that I didn't enjoy. I must admit though, NeuroSolutions is pretty flexible.

    I am not using any NN trading version, except my brain . You ought to be aware of, that I am very new to trading and nearly all of my previous work wasn't directly connected to that. I also have a notion that I am working on, however it's not really NN based (it's some type of a probabilistic state machine). I might use a NN (kohonen self-organizing maps) for a certain part of it though, like vector quantization.

    JaxPacific:
    I will forgive you, clearly, but only around 70*7 times ... and above. I hope myself .

    Now it sounds like the curse of dimensionality is something I must be afraid of, but indiors do seem redundant because there's no executable actions. Would real trading systems together with functions that create actions ? Kind upward of.FILE _ 0_EXTENSIONr and refine that information?

    By executable actions you mean buy/sell signals? It could be definitely beneficial in reducing the sound and simplify the input for the NN. But that does not mean self.

    Could you be able to memorize the information but ditch least powerful based on a specified set of parameters, while continuously accepting new info?

    That's just what a properly designed NN does. It memorizes only the most general characteristics of the training set and rejects the rest as sound.

    Understandably noted But seeing as we know the long run is and always will be just that. . Based off of what claude seemed to be going to get was more associated with probablility. I think he was saying earlier. . Referring to the likelihood of the 3 bars that are subsequent meeting buy or market states. That would not be out of the realm of an NN would it?

    It is not out of this realm if you want to do it the hard way without a actual reason. You do not require a NN for that.

    --
    Whatever you want here is to calculate the likelihood of a successful buy or market for a particular installment (correct me if I am wrong). Let's say you use two indiors for specifying installments. Those 2 indiors make up a two dimensional surface. For each point on that surface, you simulate a trade in both directions with a few predefined exit rules to find if it ends up successful or not. When finished, you use this as achievement map for future transactions. You input transactions only when the map says it's a percent of success greater than a X you choose. You get the % by finding N nearest neighbors, such as Claude said, or by definding clusters (LVQ, Kohonen nets, etc), ... or, you could be more eccentric, have some fun and do the following:

    - export the map for extended trades as an image with white pixels representing success, and black failure (or even maybe a color representing a frequency of gain, because there might be more than 1 trade per pixel, based on how you quantizated the map)
    - create two images: one for long transactions and one for short.
    - up them in Photoshop.
    - use gaussian blur with a desirable radius (that's your N-neighbours equal sort-of).
    - you can apply some additional artistic filters, but be ready for larger drawdowns ...
    - utilize the color of these pixels on the resulting blurred images as probability maps for initiating future transactions.
    - foundation money management on whitness of a pixel.


    Sounds profitable?

    --
    I wrote so many replies, I feel totally stupid now .

  4. #24
    Quote Originally Posted by ;
    hello

    Once the data was manipulated can you produce a XL file to export the newest filtered data to a different program.
    Well I know you are able to save the FFT data. This is in the form of a long list of amplitudes and frequencies. I think I have exported XY data too but if not, a great plotting program is going to have an inverse FFT function.

  5. #25
    Quote Originally Posted by ;
    Very interesting function together with your NNs .

    I have been playing with a few thoughts that could use genetic calculations myself. Basically I got the idea from reading articles about Financial Labs that uses GA to achieve some great outcomes (24 percent annum/4 percent maxDD). My research currently pertains to basically creating various approaches and then letting them loose in which the GA selects for the best performing placing most weight on this, mating successful systems and adapting to new market arrangement (mutations). You may wish to check out literature on that.

    No breakthough thoughts here (yet!) But its fairly intriguing.
    Hello Mahras

    Thanks for the advice I will certainly do some reading. Your current work appears to echo some of Dr. John F Clayburg latest work.

    There are a few big differences, however, and it's pretty contentious. He has developed a function known as. Basically the identical system runs in the background all at once with different parameters settings. The function identifies the one which is quite rewarding for current market conditions and switches over to these parameters.

    It goes against everything you have read about setting parameters for a trading system I feel the only reason he is getting anyone to listen is because of his other systems that are so rewarding. Two of them being ranked top 10 at magazine that is future cyclones stomach and system trader.

    Anyhow, if you wanna to look it over he is kind enough to provide some coding and also an explanation. You can get it here.

    http://www.clayburg.com/omw_abstract.html

    Good luck

  6. #26
    Quote Originally Posted by ;
    I know you are able to conserve the FFT data. This can be in the kind of a long list of amplitudes and frequencies. I believe I have exported XY data but if not, a good plotting program will have an FFT function.
    Thank you Mike for your answer. It looks interesting. I hope you let us know if you think of more information.

  7. #27
    Quote Originally Posted by ;
    Hello Mahras

    Thank you for the information I will definitely do some studying. Your existing work seems to echo some of Dr. John F Clayburg latest work.

    There are some big differences, however, and it's pretty controversial. He has developed a function known as parallel. Basically, the identical system conducts in the background all at one time with various parameters settings. The purpose identifies the one which is most profitable for market conditions and switches to these parameters.

    It goes against everything you have read concerning setting parameters for a trading platform I think the only reason he is getting anyone to listen is because of his other systems which are so profitable. Of them being rated top 10 at future 10, two cyclones system and belly trader.

    Anyhow, if you wanna look it he is kind enough to provide some coding and a justifiion. You may get it here.

    Http://www.clayburg.com/omw_abstract.html

    Good fortune
    This essentially creates an incredibly curve fitted model that may never replie its results. If some shift in prices occur it shall be unable to deal with it, while they can work for the short run.

  8. #28
    Quote Originally Posted by ;
    This essentially creates a very curve fitted model which can not replie its results. If any change in costs occur it shall be unable to deal with it while they may work for the short run.
    Like I said really controversial but the man is no fool it will be interesting to see what he is new systems do.They are being monitored by third party services.This is where we'll find out if it's curve fitted or gets the man actually produce something that's worth looking at.

    For a side note, I have my demonions too.

  9. #29
    Claude,

    Can it be possible to code Neural Web,Fuzzy Logic, AI on a Stage Such as VT or Even Metatrader?

  10. #30
    Quote Originally Posted by ;
    ,

    Can it be possible to code Neural Net,Fuzzy Logic, AI on a platform like VT or Even Metatrader?
    Hello amzi

    The programming which I introduced is inspired by neural networks, and shows the type of logic that they use however, is far from being in a position to replie what a dedied software would be in a position to do.

    But in case you're asking about the programming which is set up right here onnigeriaforextradingany software with the capabilities of inner loops and optimizing capabilities could reproduce this coding. Regrettably, these are just two things that VT isn't effective at yet. But Metatrader could.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
This website uses cookies
We use cookies to store session information to facilitate remembering your login information, to allow you to save website preferences, to personalise content and ads, to provide social media features and to analyse our traffic. We also share information about your use of our site with our social media, advertising and analytics partners.