The Ideal Forex Strategy Tester: Features and Key Characteristics

For all of the people who trade Forex using MT4/5 it is no mystery that the strategy tester included within these platforms simply doesn’t “cut it” for the true system development needs of a professional trader. After writing many posts about the limitations of these platforms and a post or two about our intent to build an in-house tester within Asirikuy, it is now necessary to say what the ideal characteristics of a strategy tester should be. On today’s post I will be talking to you about some of the key aspects I would expect a true serious strategy tester to have and how we expect to implement all these features within our Asirikuy testers. Through the following paragraphs I will mention why these characteristics are so important and why having them opens up the way towards the development of much more robust trading strategies.

Many people would believe that my biggest problem with MT4/5 lies in the lack of adequate Bid/Ask feed separation and the inherent lack of adequate simulation abilities for strategies which trade lower time frames. To tell you the truth developing strategies that trade in lower time frames in retail Forex using MT4/5 for trading is not only a waste of time due to the inaccurate simulations but also because of broker dependency issues which aren’t simply solvable. In my experience in order to have any level of reliability in Forex trading simulations you need to stay above or on the one hour time frame otherwise you’ll simply come up with some strategy that will vary extensively between different brokers as variations of lower time frame bars can be as high as 20-50% between brokers. It is true that if you had long term true tick Bid/Ask  data from a single broker that was truly known not to manipulate its feed you could potentially device a lower time frame strategy but such an endeavor is bound to require much more effort than the potential reward.

So if I am not interested in lower time frame simulations then what is my problem with MT4/5 ? There are several aspects of these platforms which I simply do not like and some features which are essential to the development of robust strategies which are simply not implemented within this platforms. The list is a little bit long but the main missing features that hurt good strategy development are: lack of spread changing abilities, lack of true historical spreads, lack of historically adequate conversions for quote/deposit currency mismatching tests, lack of true walk forward functionality, lack of broker dependency stress tests, lack of simultaneous multi-pair optimizations and lack of other stress tests in general.

In my view a truly useful strategy tester for Forex (even if it only uses EOD data) should have a set of tools which not only allows you to evaluate a strategy’s results with a set of parameters but which allows you to assess the degree of robustness of a given system. The current MT4/5 strategy testers simply lack any tests of robustness which is a very important part of system development. System robustness tests include variations in parameters, spread and feeds to ensure that the strategy has not been built based on false or curve fitted assumptions. An ideal tester would run several tests on slightly varied feeds in order to measure what the extent of broker dependency over a system might have been. Another important addition is the use of a true walk forward method in which the tester is capable of automatically running optimization+walk forward runs over a long term period of testing.

A key feature which is also missing in both MT4/5 is the adequate evaluation of historical swap rates and currency conversions. Whenever you’re running a simulation on a pair where the quote and deposit currency mismatch the current conversion value is used through the whole testing period for conversion. This is inadequate and generates an important source of inaccuracy in even higher time frame simulations. My idea here is that a tester should ALWAYS give the EXACT SAME results. Therefore a simulation should always be run with the same historical data for swaps and currency conversions, adequately getting rid of all the above mentioned problems.

Another requirement for a “great” strategy tester for Forex would include the ability to do “massive” optimization efforts without the need to restrict ourselves to information on just a few or even one single pair. It is certainly great to run optimizations of a single system over a wide array of different instruments, using information from all of them simultaneously (MT5 includes this but with some limitations). An important part of optimizations currently lacking in MT4/5 is the lack of flexibility in how the runs are carried out. For example if I want to optimize a variable from X to Y individually and only cross it with another variable from X+5 to Y, this is something MT4/5 simply do not allow which is easily implemented within an independent tester.

Last but not least, the statistical information presented by MT4/5 is precarious at best. A truly useful strategy tester would have a much more in-depth statistical analysis including monthly/yearly return graphs, distributions and proper Monte Carlo simulations to predict a strategy’s “worst case” scenarios. Certainly I will be working very hard to produce a reliable strategy tester to overcome many of these problems within MT4/5. Making the strategy creation process much simpler than the current process we have to face when using MT4/5. I don’t have any doubts about this being a very important step in our move towards much more powerful and accurate evaluations leading to the development of a new range of better systems for Asirikuy :o)

If you would like to learn more about my work in automated trading and how you too can get a true education in system evaluation and design please consider joining Asirikuy.com, a website filled with educational videos, trading systems, development and a sound, honest and transparent approach towards automated trading in general . I hope you enjoyed this article ! :o)

Print Friendly, PDF & Email
You can leave a response, or trackback from your own site.

6 Responses to “The Ideal Forex Strategy Tester: Features and Key Characteristics”

  1. Bruno says:

    Hello Daniel and thanks for your time,

    You are probably already aware of tonight’s unreal move in the forex market? Did anybody say Black Swan?

    Could you base tomorrow’s article on that event?

    Are your systems designed to continue normal trading under such exceptional conditions?

    Kind regards

    • admin says:

      Hello Bruno,

      Thank you for your comment :o) On the article published on March the 18th I have discussed this issue with a little bit more depth. It is however important to remeber that a “black swan” is an event of unmeasurable probability that has NEVER happened and – although rare- these big moves in JPY currency pairs within short periods of time have happened a few times during the past (therefore we can estimate their probability to some extent and a certain tackling of them is included in the modeling of our strategies). I hope this answers your questions :o)

      Best Regards,

      Daniel

  2. Franco says:

    I cannot wait Daniel,

    If I may ask, how far are your development team with the new strategy tester? Can we expect it to be available in the next year?

    I have heard of commercial testers that can do just about anything, but they are extremely expensive, ranging from $3000 to $6000.

    • admin says:

      Hello Franco,

      Thank you for your comment :o) I would expect to have a tester ready before 2011 ends, probably with most of the aforementioned characteristics. Regarding commercial testers, as you have said they are mostly very expensive – some of them are also very good – but since in Asirikuy we want to be able to share things amongst the community the best solution is to develop it from scratch (we also gain a ton of understanding along the way). Regarding commercial backtesting platforms AmiBroker probably offers the best value for your money. I hope this answers your questions :o)

      Best Regards,

      Daniel

      • Franco says:

        AmiBroker looks very interesting, not bad for $200! Will your tester be more or less the same ?

        As I see it you need at least $10 000 before you can start trading live, as the server costs will be more than your winnings if you trade less. So $200 in my opinion aint that bad if the backtesting capabilities far exceed MT.

  3. Franco says:

    Forgot to mention that http://www.tradingblox.com is one of them, any idea if they are any good?

Leave a Reply

Subscribe to RSS Feed Follow me on Twitter!
Show Buttons
Hide Buttons