Wednesday, May 22, 2013

Multi-currency EA

My last post dealt with the point and figures pattern strategy. I made some tests and optimizations and realized that this pattern is not a profitable strategy and decided to move one.
It was an interesting experiment and maybe I'll continue it in the future.
My next stop is a multi-currency strategy since I couldn't test it properly with the Metatrader as I explained in the previous posts. There is a strategy that I want to implement for some time and this is the right time for it.
This strategy took me longer than the first one. The Java libraries are very simple and comfortable as a programmer but the lack of documentation is apparent. It took me much time to find the right code samples such as  time filtering or money management.
The strategy's execution is different than Metatrader's. Instead of attaching the EA to a chart in Dukascopy you have to choose the instruments in history test, or filter the instruments in live since all the instruments signals are sent to the strategy.
My next step is performing some more optimizations, maybe insert some improvements such as trailing stop loss and forward tests.

Thursday, April 18, 2013

First steps in JForex - point and figures expert advisor

Except the obvious advantage of this platform which is the real tick data it has a few more advantages over Metatrader 4. One of them is the the chart types. The first chart that drew my attention was the point & figures chart type.
There are solutions for Metatrader but those are far from being perfect. The main problem is the accuracy and the second problem is that it's not a native chart.
My first goal was implementing a point and figures pattern and test it with real tick data. This will allow me to experience the new platform, the API and the test mechanism.
I spent a few hours until my mission was accomplished. The platform is JAVA based. There is an option to connect to it through Eclipse but I couldn't log-in. In the future I'll get back to it and solve it with their support. It looks like a great interface with the capability to debug with a real debugger.
The API requires programming knowledge. It's a little more complex than MQL 4.
There some issues that I found the answers by digging in the API, looking in Dukascopy's programming forum and some questions remained unanswered. There are much more sources for MQL4 / Metatrader in the net than JForex and that's the first disadvantage I found.
The Metatarder's GUI is more convenient.
The EA itself works fine. I'd like to verify that real test results are the same as the backtest's and my next stop is performing an optimization.

Sunday, April 7, 2013

JForex

It the past few months I had a lot of work and couldn't dedicate much time to my private research.
Recently I was asked by one of my customers fro a small project in JForex. For those of you who are not familiar with it I'll explain: Dukascopy has a web trading interface with Java based API. This is the programmable gateway to automated trading in this platform.
At a first glance it looks simple compared to the Metatrader. I checked in the past and moved on becuase it looked very primitive to the tools I was used to.
Each time I investigated the options to setup a reliable back test system I found that Dukascopy is a good and  free source for reliable historical tick data.
This is my short background with JForex until that small project I had last week.

Well, now I'm in love!

Why?
I'll explain:

  1. The platform is completely web-based. This way the data is not stored locally on your machine and it ensures the same historical data for all the users.
  2. The programming interface is done with Java. Java is like English... very common and almost  everybody speaks this language in the programming world. For me as a programmer it allows me to write fully object oriented expert advisors. From that point of view it is clearly better than MQL4. MQL5 is a different opera but it's not as common as MQL4.
  3. I think that this is the feature that hit me: the backtest. This platform allows the user back test his trading robots with real historical tick data. I think that the biggest Metatarader's flaw is the historical data and back test mechanism. First, the historical data is not accurate most of time. Different machines with the same platform has different history data. Second, there are no real ticks in historical data. The highest time resolution is one minute and then there are tick modelling techniques. This is a mandatory feature for automated trading.
  4. Multi-currency back test - this is something that did not exist in Metatrader 4.
Combing all of these together allow me to fully develop and test multi-currency trading ideas that showed great potential a few months ago. The key feature here is 100% accuracy. I don't claim that the past results ensure me future results but accurate past test is crucial.