Advice on not using the MetaTrader 4 Strategy Tester - page 7
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Those traders who have spent a fair amount of time testing have already done their tests and are satisfied with the quality of testing and the margin of error. It's a pity not many people publish their tests.
I did a comparative analysis of testing in the tester and online trading a year ago, when MQL4 appeared. It would be boring to write an article about it, but I can certify that the level of tick-flow modelling in the tester is 100% satisfactory to me. I have written more than 30 Expert Advisors of various concepts and a lot of variants of them during a year or so. At the beginning I meticulously checked results of testing against on-line trading literally for every tick. The difference in two or three pips is not significant. It is the same as the situation when one sets a 50 point profit and regrets that the market did not reach the profit level by 2 pips. Or a stop loss caught on the very edge of the market has worked. It worked, so be happy with it. This does not affect the conception of the trading strategy. Catching fleas is a waste of time.
At the same time, the tester very accurately selects the first tick of the bar, which is fundamentally important, and quite accurately ticks the nature of the market within the bar, which is also nice.
The tester is a super program, having a normal minute history and knowing how to use it correctly (and running normal EAs on it)
You can get results almost 100% similar to real trading. I have checked it many times.
Yes, but, sorry, I find the difference of 10% between real and tester to be INCRITICAL. This means that the minimum expectation should be 10!!!! It's one thing to say, don't writereal with 0.25 expected payoff, and quite another to say, write only with 10!
All this is even multiplied by the fact that slippage may occur by 2!...points (and in both directions). So, instead of understanding the real tick history, we have been collecting statistics for a month already and instead of doing our job :)))))
What mode are you using in the tester?
i only use "by opening prices (fast method on completed bars)", no interpolation
And of course I download all the minutes
Here the Expert Advisor is working on the Championship (if we don't count the first two trades we had due to the lack of history at the moment of start)
Yesterday's day and part of today's day passed - I ran the Expert Advisor in the Strategy Tester, all operations coincided
I only use "Open prices (fast method on formed bars)", no interpolation
And of course I download all of the minutes
But I only use the "All ticks" mode. Firstly, I always have a powerful computer at hand, and secondly, to speed up the calculations, I first make the step of optimized parameters large, and then gradually decrease the step to "1" and at the same time narrowing the range of optimized parameters.
I, on the other hand, only use the "All ticks" mode. Firstly, I always have a powerful computer at hand, and secondly, to speed up the calculations, I first make the step of optimized parameters large, and then gradually reduce the step to "1" and at the same time, narrowing the interval of values of optimized parameters.
similarly :)
I haven't yet had time to benefit from tester visualisation mode, but there are thoughts of using it effectively.
Stages:
1. Optimise the Expert in the tester.
2. Using the Expert Advisor in the online mode on a demo account.
3. Obtain the Expert Advisor operation results on-line including detection of its bottlenecks.
4. We re-run the Expert Advisor in the tester using now visualization mode on historical data for the period of its operation in online mode. This helps to recreate the emotional component of the Expert Advisor's work. Analyze, work out the errors, correct the Expert Advisor.
5. Repeat steps 1-4 until an acceptable result is obtained.
All the same, psychologically and mathematically it is quite different - ticks, modeled by the tester and the real. Because this is a subject for long-term considerations - what contribution into real trading will the replacement of simulated ticks by real ones bring - and it is far from obvious that an EA with a BIG expected payoff will not accept the replacement - and what if the tester missed a stop loss by half of your deposit, while the real one did? Any expectation flies in the face, because:
a) There may be too few of such shortcomings on the history to statistically evaluate their significance.
b) It is not clear how to catch them on the history - in fact on real ticks you cannot run them. It is a vicious circle.
All the more, that this trick is not super difficult to implement, I hope that in future versions/builds we will see it.