My EA is giving me 90% wins but I have an issue

To add comments, please log in or register
g_r_a_robinson
6
g_r_a_robinson  

I have an EA that gives me over 90% wins on longs and shorts on any currency pair and on any time frame. Which is acceptable to me. However I had been backtesting only using the :-

Control points (a very crude methods based on the nearest less timeframe)

When I switched it to :-

Every tick (the most precise method based on all available least timeframes)

It just fails miserably. I realise that even the 'Every tick' method is never 100% yet I also know its more accurate the 'Control points' method. I'm quite new to MQL4 but I've determined that the only way I can make this useful is to find out how to interpolate the timeframes the same way that the strategy tester does and apply it in my code.

I wish to use the H1 for live testing. So the question I have is what least timeframe would the H1 strategy tester be using?

I counted the number of times the bar moved on the H1 and it appeared to be 12 times, which lead me to think it was using the control points of the M5 bars. If this is the case how on my H1 chart can my EA access the time period for the M5 time period? I thought about having a separate M5 chart open and saving the results to a file that my H1 chart can pick up and use. But there must be a better way of doing this.

Does anyone have any ideas on this one?

Thanks

George

MGTeam
4
MGTeam  

Hi George,

You and everbody has to know about one of the classic scams when you buy an EA and see the wonderful results. The first thing to have a look onto is the method that you have described.

CONTROL POINTS backtesting method is a very rough way of testing. The tester modul jumps on the next upcoming bar, and simulates some intra-bar data and sends it to the tested Expert Advisor. These points are NOT based on real M1 or TICK data, it is only a basic-basic, almost randomly simulated data inside the current Bar. When you test it on higher timeframes (M30, H1, H4, D1, W1) it certainly gives some type of result, but you must know that it doesn't count at all, it is not realted to the reality. Most EA programmers know this, and they are able to optimise their system to show beautiful result with Control Point Bactest Method.

EVERY TICK is a much more realistic method of testing, however it does not cover the real moves 100%. The MT4 has a built-in short term "memory" of the recent ticks, and has got some sample files of ticks from the past. So, by choosing this method, when the Tester jumps onto the next H1 bar (or higher timeframes), first it "unfolds" or "decomposes" the H1 Bar into M1 bars (if this data is available, which I highly recommend to get). If you haven't got the "high-definition" data of M1, the tester will use M5 and so on.

Inside the M1 bars the tester uses the described Tick data samples to simulate the intra-minute moves of the currency.

This method gives realistic results on strategies that tries to capture medium to bigger profits of pips.

For a strategy that hold a position only for seconds or 1-2 minutes, or tries to get smaller profits than two times (2x) the average minute range (only 3-10 pips), this method ALSO FAILS to give realistic results.

Sorry for my imperfect english, I hope this helps.

To add comments, please log in or register