Backtesting Problems

Back to topics list To post a new topic, please log in or register
ernest02 2016.01.12 14:30

I have written an EA that is quite sensitive to small price changes of 2 pips and more. When testing on historical data in the Strategy Tester it gives beautiful results, but when doing forward testing it is very disappointing and not profitable.

When I retest the EA on the live data used for forward testing - which has now become historical data - I get wonderful results much better than when doing live forward testing on that same data.

It therefore seems that the data for back testing has been changed from what it was when it was reported live in MT4. Is there a solution where I can get the historical data in the same format when it was reported as live data so that I can do back testing that can be used successfully in forward testing with live data? I think I may be talking about the original tick data generated by a live system.

Would be very grateful if somebody can come with a solution.

gooly 2016.01.12 15:54 #

Do you know that mt4 creates its own ticks acc. to a unknown algorithm?

So it can be that you have adapted your EA to the algorithm but not to real data?

ernest02 2016.01.12 16:51 #
Yes. I discovered that the MT4 Strategy Tester uses 1 minute data with H,L,O,C values and then generates the ticks randomly. So any test with this data with a EA who uses small increments (e.g. scalping) cannot be accurate. So I will have to download real tick data from a broker and then convert it into FXT and HIST files to use in MT4 ST. A BIG job!
gooly 2016.01.12 17:21 #

... and then generates the ticks randomly.


So I will have to download real tick data from a broker and then convert it into FXT and HIST files to use in MT4 ST. A BIG job!

1) I doubt "randomly"! There must be an algorithm otherwise every new run of the Strategy tester would end in difference results - which nobody would accept!

2) At least you (mt4) can backtest against real tickdata while mt5 can't!

ernest02 2016.01.12 21:14 #

You are right Gooly! The ticks cannot be random. At least they should have a direction from low to high e.g. if the Close price is higher than the Open price - and vice versa.

I have downloaded some tick data, converted it and uploaded it, but still the results with the Strategy Tester looks much better than my forward testing for the same period.It seems there is more to this. :-(

ffoorr 2016.01.13 10:20 #

A scalper which open or close order  taking very small profit,  can be backtested on MT4, but the result will not be reliable.

At the moment of high impact news release, the scalper will open an close a lot of order on the one minute timeframe; Making a lot of profit  -- But this is possible only on backtest.  In real trading there is slippage and requote, especialy on news release. At this time, it is not possible to open or close order at any desired price. This is why a lot of people complain about scalpers on the net, i guess.

This assertion is true if the expert open and close orders in the same bar, otherwise if orders are opened on one bar and closed on another bar, it may be possible to make money and backtest scalper, (I did not try to do it, but seem plausible)

EA Scalper can be found on the Net, they analyse the market, and when high volatility market conditions are met, when 1 minute bar size are over 200 point to  say 400 point, they open order. This is just the condition encountered at news release time. With good money management these EA make marvels (wonders?) on backtest.

ernest, i am afraid that looking for tick history won't solve the problem, even if taking small profit with good money management is a good strategy.

Back to topics list  

To add comments, please log in or register