Nine stages of backtesting of Forex trading system with the help of Forex trading indicators

1. Development of a hypothesis. Start with considering behaviour of the Forex market and, proceeding from careful supervision and laws of logic, build a hypothesis (Forex trading strategy).

2. Data gathering. Load as much as possible the historical data on the financial tool interesting you. The abundance of the processed information provides the big statistical importance of the received results.

3. Check of the data. Carefully having checked up the data, make sure that does not contain in them of gross blunders. The elementary way of check - construction of schedules on which emissions and uncharacteristic movements are always well visible. The received data should be compared to the information from independent sources. The qualitative data - a necessary condition of productive backtesting of Forex trading indicators and Forex trading systems.

4. Segmentation of the data. Divide a database into intervals - years, quarters, months, weeks. The first segment make larger, than other.

5. Optimization. Study the first segment and allocate the specific parameters, capable to increase efficiency of a ratio compensation / risk in this time period. As optimization understand regular search of the best parameter of the Forex trading indicator - the best, that is giving the highest and constant profit at the least falling the capital. The optimum parameter for the first segment is looked for by simple selection: in the chosen formula of the Forex trading indicator the size of parameter varies, and its value at which the size of the profit concerning risk (the maximal falling of the capital) appears the greatest thus is looked for. We shall remind, that optimization should be carried out extremely in limits of the earliest on time of a segment of the data. This first segment used for search of the best specific parameter, it is accepted to name the data within the limits of sample; The data of later time, till now not exposed to the analysis, are called as the data outside of limits of sample. Earliest of the data outside of limits of sample in due time will be used for continuation of check (a stage 6). At a stage 5 it is necessary to allocate the specific parameter connected to an optimum ratio compensation / risk, however to fix the information on efficiency, proceeding from the data within the limits of sample, does not follow. The results describing efficiency of parameter, can be counted up only after carrying out of the analysis of the data outside of sample (that is at a stage 9).

6. Continuation of testing. The found optimum parameter of the Forex trading indicator (received on a stage 5) is checked on the following nearest by the present moment, a time segment of the data which were not used during selection of parameter, outside of limits of sample. The data on efficiency are fixed for the further calculation of results (a stage 9).

7. Addition. Combine a segment used for check of parameter (at a stage 6) with the earliest segment (a stage 5). After each such operation the quantity of the data in sample should increase, outside of sample - accordingly, to be reduced.

8. Repeated recurrence of the procedure which have been lead at stages 5, 6 and 7, stopping with disappearance of group of the data outside of sample. Begin the process of optimization described in item{point} 5, using the data, just (after a stage 6) including in sample, - the quantity{amount} of the data in sample this time will be more, rather than at a stage 5. Having repeated optimization (see a stage 5), check up the received parameter on the following segment of the data outside of sample (see a stage 6). At last and this segment (namely with what we worked in item{point} 6) joins the general database (from item 5). Stages 5, 6 and 7 repeat until the quantity of the data outside of sample will not be exhausted also we shall not come to the segment of the data corresponding to the present moment. Similar operation can seem labour-consuming, however in comparison with check in real time when for a day the behaviour of the Forex trading indicator within one day is exposed to backtesting, it looks fast and easy.

9. An estimation of results. If the analyzed data not accidental, and the quantity of segments is great enough, during backtesting you receive worthy trust the information on behaviour and efficiency of the chosen Forex trading indicator for a number of years. You have found out, as far as profitable and stable there was a Forex trading indicator, and now have an opportunity from several Forex trading indicators past check to choose one for the most successful work in real time. Your hypothesis, thus, has passed objective check by results of which it can be either is accepted, or rejected.

In a case if the parameter tested on a material of all data set outside of sample, appeared effective, it is necessary to lead last optimization on the basis of all available data and to receive value of parameter for subsequent use in real time. However it is not necessary to pass nine stages described above and at once to pass to the test on the basis of all available data as thus you do not receive the major information about weak and strengths of the Forex trading indicator, no less than about evolution which it undergoes eventually.

Carrying out modelling, you can leave in sample all seen data irrespective of as far as they have become outdated. Laws of the Forex market are established by people, and the human nature varies with current of years a little. The old data, thus, can be rather significant, and behavioural models, characteristic for the past, - to tend to recurrence.

And on the contrary: for acceleration of process is allowable to delete regularly from sample a quantity of the data concerning by the earliest period, replacing with their equal quantity of the data from a segment outside of sample. Differently, for revealing specific parameter " the sliding window " established length can be used. If to admit, that the basic characteristics of the Forex market essentially vary eventually, aspiration the analytics regularly to lower the most "ancient" historical data it is represented proved.

The process of testing described by us represents the Forex trading analysis as a method possessing a high degree of scientific accuracy. In a case if received on a material of the data in sample the parameter yields good results with reference to the data outside of sample, use of the Forex trading indicator in the near future, obviously, in a high degree is justified.

Undoubtedly, to be convinced of suitability or an inefficiency of the Forex trading indicator during the analysis of the historical data where it is more pleasant, than to receive the similar information in real time and with real money. The investor wins irrespective of, true or erroneous there was a hypothesis. In opinion of Thomas Edison, the rejected theory is the most useful information as allows to concentrate attention to the phenomena and methods which studying can give positive result. Process of modelling, allowing to look at behaviour of the Forex market more steadfastly, is a source of numerous new ideas.

Copyright 2007 Gordago. All rights reserved.
Copying of all making parts of a site in any form without the permission of holders of copyrights is forbidden.
EnglishEnglish RussianRussian