Is MT4 tester reliable

New comment
 

Hello all,

I tested my simple EA on MT4 startegy tester. I noticed that my EA is too good to be true so I decided to look at details and here what I found:

My EA tested on H4 period wth EVERY TICK (best) option.

1. EA sent SELL STOP order at 1.6116 on 2010.01.04 00:00 am with SL 1.6196 and TP 1.6035

2. The order was filled on the same candle (period)

3. Tester closed an order with take profit next day i.e. 2010.01.05 8:00 am

Very fine but the history records shows that about 11:25 pm 2010.01.04 the qoute was 1.6238 and the SL should be fired. You can see this below

I show here 4 hour period of history but I checked also 1,5, 15,60 min candles in history and all "say" the same: about 11:25 4 Jan 2010 the price was 1.6238 ! and SL should be triggered.

I want to say that I used clean MT4 with my own history qoutes which I had imported from forexite.com. I also block MT4 from automatic downloading any data from broker or metaqoutes (just I didn't allow to connnect to internet).

This is only one example but I saw the same behavior before many times in different situation. I'd be obliged to hear how to explain this behaviour.

Andrzej

 
Andrzey:

This is only one example but I saw the same behavior before many times in different situation. I'd be obliged to hear how to explain this behaviour.

The quotes in history are Bid prices, but Sell closes at Ask... Please check the value of spread before Testing and verify it makes sense and is smaller than the diff between the 2 prices u quoted (if it's larger, then it shouldn't have closed).
 
MT4 Tester is good for testing the strategy, ins & outs from the market at the right signals, in terms of reliability for profit, it is very buggy. Forward test your EA for a month on a minute data, then run backtesting for that same month, compare yr results. When you compare, If you have exact entries and exact profits between backtesting and forwardtesting, then you can assume that for higher time frames your results will be similar, and that your EA Tester results can be used with good probability.
 
c0d3:
[...] in terms of reliability for profit, it is very buggy.

It's not. The limitations of the Tester are well documented (https://www.mql5.com/en/articles/1512). The problem is that most users don't know these limitations... It's not buggy, it's just limited in accuracy.


[...] When you compare, If you have exact entries and exact profits between backtesting and forwardtesting [...]

You would never get 'exactly' the same, because of the limitations mentioned before, but even if the Tester was perfect and had no limitations, there would still be differences caused by some random aspects of live trading which are impossible to replicate (for example how would the Tester know about latency issues, cases of server overload, disconnections, etc. -> all these affect trading).

 

Adding to gordon's list of real-world trading environment characteristics that strategy tester fails to capture/replicate in backtesting is the variable spread and slippage/order-requotes.

http://trading-gurus.com/the-metatrader-4-backtesting-blues-track-1/

 
1005phillip wrote >>

Adding to gordon's list of real-world trading environment characteristics that strategy tester fails to capture/replicate in backtesting is the variable spread and slippage/order-requotes.

/go?link=http://trading-gurus.com/the-metatrader-4-backtesting-blues-track-1//

 

I rewrite my EA to work in 1 min periods as cod3 said. It causes some problems with statistical calculations etc but it's possible to do. Now I compare it with my own tester writen in c# and both give allmost the same results. (there are some differences because they don't calulate statistical values indentically). Thank you guys for quick answer. All in all I get satisfied with testing and can trust the EA.

New comment