Log in

Shortcut to Discovery Computerized Trading System Development and Modeling by Mark Brown


This is a very interesting presentation about system trading and development. The theory of the switches between market states. However he distinguishes just two major market states.

So his idea is that there should be an algorythm to switch between the market states:

  • If such and such then switch 1 = 1
  • If such and such then switch 2 = 1
  • If switch 1 + switch 2 = 2 then do this

And that follows that you have several systems for different market states and combining them you achieve an optimum performance:

  • Non-Trending entry method switch
  • Trending entry method switch
  • Combining the entry method switches
This kind of ideas is a recurring topic among professional traders.


  • JohnLast 3969 days ago

    To develop a range trading system is not that easy. I think machine learning can help a lot to develop a succesful range trading system. 

    However it is not easy because your range trading system may loose, when your trading system is performing very well and it may be better to let your trading system to do the job alone. 

    And vice versa your developped ranging trading system may perform and compensate the loosers of the trend system.

    The question is to be positive when you combine both equities.

    It is a classic to develop a ranging system with low targets and large stops. In that case it is advised to trade only in one direction - the direction of the prevailing trend. The Euro Swissy is a classic example of that, using a stop based on an important support level.

    So developping a succesful range trading systems is not that easy, many things need to be taken in consideration.





  • tovim 3969 days ago

    If I remember correctly this guy http://championship.mql5.com/2011/en/users/Tim/discussion   managed to create the market state regime with zigzag and neural network.

  • JohnLast 3969 days ago

    This is just another sophisticated system with suicide orders, specially designed and optimized for demo competitions.

    If you look at the percent of winners and the equity, it is obvious that it is using very big stops.

    And it is closing to the zero the orders that do not go into the right direction.

    This does not mean that this would not work in practice but the money management would be completely different for a real account

    As for the zig-zag and neural networks this is the standard approach for trading solutions.

    They apply a kind of zig-zag on historical prices and they call that optimum trading signals. Then with neural nets they try to model the optimum trading signal. And yes it works. 

    But it is not sure that Tim did that in his EA. Maybe he was just looking at the direction of the Zig-Zag and then waiting for a break - out with an order for determined time. That explains the number of the failed orders.

    And with an EA he may have optimized the time and the parameters of the Zig-Zag.





  • JohnLast 3969 days ago

    Maybe you should take a look at an interesting AI project using pending orders.


  • tovim 3967 days ago

    Thanks John for the link.

  • JohnLast 3967 days ago

    Mark Brown would be glad. I wrote him and thanked him for the article and that we find it most interesting.