Walking Forward: Do Asirikuy Systems Yield Profit With Continuous Optimizations ?

A while ago I wrote a post about how I didn’t believe in short term optimization and why I believed this sort of mechanism does not work for systems in the long term as you are continuously “chasing” market conditions. However it is true that a system which does hold up against “walking forward” within a long series of tests is robust in the sense that there is a strict technique for the adapting of the system against developing market conditions. In order to make sure that a system does yield profitable results in this way it becomes very important to evaluate the effect of the technique across a very long period. Within this post I will describe the results of some Asirikuy systems in this regard and how this opens up a new route to system adaptation.

Let me now explain to you what the walking forward technique is all about when it pertain to short term optimizations. What we do is simply take a system and optimize it for a period of days in the past (for example 120) and then run the system for a given number of days (for example 30) and compare the results of both tests. The walk forward period is generally about 20-30% of the optimization period since we assume that optimized systems “become old” quite quickly.

Now the idea is not simply to run one of these tests but to cover a very large trading period – for example 8 to 10 years – using this exact technique. What we do is run tests for every 30 day period within the 8-10 year test using the optimized settings for the past 120 days. If a system is able to survive through a ten year period after this process then it shows that the system is able to survive to the “walk forward” and therefore the strategy is robust in the sense that short term optimizations do generate settings that are generally profitable through the out-of-sample period. The process to do this whole evaluation is extremely slow (since each 30 day period requires an optimization of many parameters) and then a walk forward. In summary – depending on the depth of the optimizations – it may take hours for each “30 day step” for a 1 hour system.

Although someone already thought about this problem and developed a program to perform this sort of walk forward analysis using MT4. The analysis tool – found here – makes the analysis of systems in a walk forward manner much easier although it suffers from great limitations which make its practical usage almost null since the software is very limited as it can only run optimizations and runs with the same backtesting technique something which makes “every tick” runs impractical and the overall methodology restricted to systems that can give reliable simulations based on “open prices only” (systems that do not rely on either Close, High or Low values).

The best solution – which is what I have done – is to develop a piece of software which allows the running of a walk forward analysis by optimizing a system using a method based only on OHLC values of the time frame were it trades (very fast) and then run the resulting 30 day result using a model based on 1 minute OHLC values (analogous to the “every tick” method in MT4). In the end a solution like this based on FreePascal is able to run the above tests in a very fast way and to cover a 10 year period with optimizations every 30 days in less than a few hours, with the final runs being done on a higher modeling quality than the optimizations (which if done at a higher quality would take much longer).

Up until now I have limited my analysis to systems which I can easily code – such as Teyacanani and Watukushay FE – but the results are pretty good in the sense that they show that Asirikuy systems – when run this way – give a profitability which is in average between 40-60% that of the previous optimization period. In this sense Asirikuy trading systems are robust since they show their ability to survive to constant optimizations against the market through long periods of trading. Although chasing the markets, the systems are able to succeed.

The obvious question now is whether or not this is better than using a single set of parameters through a very long period of time (the way in which we’ve been working). The answer to this question is not very obvious or easily obtainable since we currently lack the necessary out of sample data length to know the profitability of systems within an out-of-sample test of 10 years but from what I could see whenever the optimization period grows so does the matching of the system’s profitability with the period. For example while a 120 period optimization with a 30 walk forward generated an average profit match of just 40-60%,  increasing the optimization period to 1200 days and running the system for the next 300 yields results in the region of 60-80%. Perhaps the best way to develop systems is then to have profitable parameters developed for as long as there is data available and to then re-optimize the parameters when 20% of that data has gone through. In our case this would mean that re-optimizations of Asirikuy systems would have to be done first in 2 years and then in longer periods as the “testing body” grows bigger.

Certainly the notion of walking forward systems is interesting but from the information I have been able to gather it seems conclusive that short term optimizations simply try to “chase the market” and long optimization periods are required to improve the results of the out of sample periods. Certainly an analysis of more systems on more currency pairs is required to validate these conclusions but up until now they seem to hold up :o)

If you would like to learn more about my work in automated trading and how you too can learn how to build your own likely long term profitable systems based on sound strategies 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.

3 Responses to “Walking Forward: Do Asirikuy Systems Yield Profit With Continuous Optimizations ?”

  1. Jacob says:

    sounds good, daniel:-)
    does yor tester include an option to randomly “scatter” the testing intervals (the ones for optimization) and not just, say, from jan-dec ’02?

    • admin says:

      Hello Jacob,

      Thank you for your comment :o) The program does an X day optimization followed by a Y day walk forward. The optimization can start anytime and the walk forward will always follow it (as this is the intend of the test -> forward). So what I did wasa ten year test which was a compilation of each one of this runs. So for example I would run the optimization from Jan to Apr 2000, then test May, then to generate Jun I would reoptimize from Feb to May and test Jun doing this over and over again until I achieved the 10 year test. However since the optimization and walk forward are a fixed number of days (not months) the values shift and you never evaluate strictly a “set month period”. I hope this answers your question :o)

      Best Regards,


  2. Vitor Ramirez says:

    Hi, Daniel It seems interesting your site. In a market full of scams, it`s easy to know when a program is serious Most probably I will join your site soon.


Leave a Reply

internal_server_error <![CDATA[WordPress &rsaquo; Error]]> 500