If you think that the data, obtained during testing (average annual profit, etc.) in the Performance tab are true, then, in most cases you will be wrong. And it's just a little thing, which was in the fourth version, but which was gone in the fifth and sixth — Lead Bars.
For example, want to test the system on a daily timeframe for a year from 1 января до 31 December. And the system uses MA with a period 250. And if you test the system in a year, you will not get a single deal, since MA is not formed yet, there will be no deals, and it will be formed only after 250 Bars, which is exactly one year. That is, the profitability for the year and all other indicators will be in a highly distorted form.. In this case, just get all zeros. Period 250 I gave it as a last resort. Perhaps the system will have indicators with a shorter period, for example 50 or 75. But in this case too , testing will not start from the beginning of the year, and after three to four months, when indicators form. AND, respectively, all system indicators will be distorted because the test was not for a full year, and in eight months, for example. And the tester claims that it was in a year.
If about this, contact support, then you will be advised to start testing on the wrong date, with which you want to test, and from date, which will be obtained by subtracting a certain number of bars. For example, as in the above case, not start testing with 1 January, а отнять 250 bars and start testing with 1 January last year. Then the indicator with the period 250 will just be formed on the date you need, i.e 1 January this year. Everything seems to be fine — all signals will go on time and capture all trades throughout the year. But, in this case, the bottom line will be that all the data of your tests will already be calculated IN TWO YEARS. I.e, for example, during the test you will receive 30% per annum, but in fact, the program scatters this profitability over two years, since we actually tested it for two years. And the average annual return will no longer be 30%, and 15%. And other data and coefficients will also be distorted..
Certainly, if test, for example, for a very long time, for example years over 40 and an additional year to allocate for the formation of indicators, then the distortion will be almost imperceptible.
Another factor, Equity vs.Buy-and-Hold Curve" — it just won't be able to use it correctly, since buy-and-hold" will always start from the actual test start date. And the test equity from the beginning of the formed indicators. I.e, in the above case, their start will be given with a difference of 250 days (Bars).
Certainly, everyone already understands that historical tests remain only tests for an approximate estimate of the health of systems and past data will never be repeated in the future. Approximately this is how the WLD developers explain this incident on their forum in response to customer requirements to reconsider their attitude to the issue., flatly refusing to introduce Lead Bars, as it was in the fourth version. Probably, for this they would have to reprogram a lot of what is already programmed — otherwise their obstinacy cannot be explained. But still, I think, this is not entirely correct.
This problem is unique to WLD 5-6. Other programs do not have this problem.
