Friday, September 3, 2010

Why are Five Years Statistically Significant for the Evaluation of Trading Systems ?

Every time someone asks me what I consider to be the minimum necessary period to evaluate a trading system to know if it has a chance of being profitable under future market conditions I unequivocally say "five years". However - although I have explained vaguely in the past why - I have never written a precise explanation that tells people exactly why this is the case and why systems evaluated over at least 5 years have a better chance of surviving than those evaluated over a 1 or a 2 year period. Through the following paragraphs I will tell you the reasons why this is the case and why using this as a minimum period of evaluation guarantees that systems will have a certain degree of adaptability and the possibility to survive to future market conditions.

The first thing we need to ask to understand the 5 years argument is : What changes when market conditions change ? When you analyze any trading instrument and look for changes in the different quantitative characteristics of the market you will notice that changes in market conditions are usually accompanied by changes in volume. This happens mainly because market participants trade more under rough market conditions and less and more orderly under growing market conditions. In the end what you have in an economic cycle is a series of cycles in volume. Since volume is proportional volatility (which is just a way to measure the length of the movements within an instrument) we find that volatility changes as market conditions change. Generally markets in which the economy is growing are steady and quite non volatile while markets where there is a lot of economic turmoil are extremely volatile.

If we then consider that changes in market conditions correspond to changes in volatility then in order to have a sufficient variety of market conditions for the evaluation of a trading strategy we need to have a large enough amount of change in longer term volatility. When you look at the daily or weekly charts of any given instrument, you will notice that volatility within most years tends to change very little while periods of 5 years usually contain large changes in volatility. The graph shown below of the EUR/USD weekly chart clearly shows you the changes in volatility during the past 10 years (as the 14 period ATR indicator). You can see how any given one year period has an almost static volatility while periods of several years, especially 5, have large changes in volatility.
-
-
The period of 5 years comes from an analysis about these variations in volatility. If we take a look at any instrument and consider the time it takes for an instrument to go from its average level of volatility to a new high and a new low and return to the original level we find that this period is roughly 5 years (like how it is shown above). This means that after a period of five years there is a large amount of different market conditions that a system needs to tackle if it wants to be successful. Therefore a system that survives to testing periods of more than 5 years has a high like hood of surviving to changes in market conditions in the future since it contains - within itself - the capability to adapt to changes in market conditions.

Of course, a 5 years period does not implicitly guarantee that any given system will be able to achieve success in the future since the market can change further or at a faster phase than what the system sustained during that 5 year testing period. However it is true that to survive profitably through such a long period a system needs to have some degree of adaptability that is not necessary to survive to shorter testing periods when hardly any changes in volatility happen during most years. It is for this reason that evaluation of strategies through prolonged periods of time is necessary since short tests of just a few years may only show how the system behaves under some very specific market scenarios. Of course, the longer the period you use for your tests and the larger the overall chances in volatility, the more robust your system will need to be.

If you would like to learn more about the evaluation of automated trading systems and how you can evaluate and create your own systems that return profits after 10 years of evaluation without exploiting any backtesting faults please consider joining Asirikuy.com, a website filled with educational videos, trading systems, development and a sound, honest and transparent approach automated trading in general . I hope you enjoyed this article ! :o)

No comments:

Daniel Fernandez's Expert Author Email Alerts
Sign up to receive email alerts of Daniel Fernandez’s latest articles from EzineArticles.com!

Email Address: