Testing adviser for beginners


There are different approaches to testing advisers. Every trader, probably developed his own approach to testing of mechanical trading systems. This article will be useful for beginners who have not yet formed an opinion about the methods of testing.


During the development of mechanical trading system, as a rule, optimization of the system, the selection of parameters, the addition of various conditions and filters. This is a common practice among traders use it mechanical trading systems. Then, necessarily, being tested adviser.


The most common method among traders - beginners are tested for all of the available history of the quotations. Because it is believed that the greater the gap, which tested the system, the better. Optimizing the system available to traders throughout history, he thereby depriving themselves of the opportunity to test the so-called «out-of-sample», ie when the system is run on historical data, which did not participate in the optimization. The check on the performance of the system in this case, the trader must be carried out on a demo account in real time.


In this method has its pluses and minuses. Of plus you can highlight what is used to optimize the greatest number of bars of available stories. And the more time the expert testing, the more he will make transactions over this period. Well, the more transactions, the higher the reliability of the results. Minus only one - the inability of the traders test the system on historical data, which did not participate in the optimization.


Perhaps, many experienced traders say that the test «out-of-sample» need for absolute certainty, to hold on a demo account in real time, and will be right. But such testing would require a lot of time - several months or even years, for beginners, as indeed, for the trader with experience, it's just not realistic. And yet, after such testing, the trader may receive a negative result, and time lost. To avoid this, it is better to use the second method of testing.


To use the second method of testing the system, a trader should break all the available history on two segments - long and short. On the long stretch of being optimization and tuning of the system. Over short distances, testing is done "out-of-sample". Long plot contains about 80-90% of all available quotes history. Each trader decides for himself which part of the history and what to use.


The principle of testing is very simple - first, on a long section of optimization of the system. After receiving the results, we select the most reliable (how to choose them, you can read the article "What are the optimization results to choose) and start testing the system on the same long stretch. Thus, we obtain the values of all system indicators - profit, profit factor, drawdown, the number of transactions, etc., which actually describe the behavior of the system in history.


Now, drives the same system on the second - a short interval, which was not involved in setting up the system (in fact this test is called "out-of-sample"). Having tested the system in such a way, the trader will receive the new service indicators - profit, profit factor, subsidence, etc.


All that remains now is to compare, how the system behaved in a long section where there is a configuration system, and how it behaves in a short interval for testing "out-of-sample". The system should show similar results. If at least one parameter of the system is much worse, then we can assume that it is non-operating or perform additional tests for authenticity.


Many traders complain that the comparison of test results, which were conducted at intervals of different length of history is not quite correct. So it is, because the long part of the story contains a different species or types of market: downward trend, upward trend, Flat, etc., while the short part of "out-of-sample", most likely contain only one of these types.


In order to properly compare the results of a trader to choose from most of the area equal to the period of the test «out-of-sample» and compare these results. The system will be operational, only if the test results «out-of-sample» are no worse than the results of any of the selected trader periods.


The trader can always conduct additional, more sophisticated studies to confirm or refute the system work. But at an early enough stage of testing described above, the more it is easy to use and the trader at any level will be able to use it.


One important tip in the end:

It would be a good trader does not test the system on the history, it can never be 100% sure that the real market its system will behave as well as in testing. Therefore, always, before you start trading on the system, the trader must define the criteria for stopping trading on it. The most simple criterion - drawdown, if the system to the realities of a possible drawdown greater than during the tests, you must immediately stop the trade. This simple rule will allow novice trader to avoid losing your deposit.

"Forex Never Lose Trade"