Any way to reliably test expert advisors?

 

Hi,

i spend a lot of time searching and testing ea's, i even started to learn mql5 to develop my own ea's.

But what i learned by now is that there is no reliable testing method for expert advisors - strategy tester does not behave like a real account and even the demo accounts i tried so far are not really reliable.

Example?

Look at  https://www.mql5.com/en/signals/15563 which was a test of a expert advisor i have developed. More than 50% growth within one and a half hour - not bad, right? 

Well, this ea is a kind of fake - it only works on accounts which are NOT using market execution and on a real account it does not trade at all because of requotes.

 

So, how can i test if a expert advisor will be profitable if even demo accounts are not reliable? 

Trading signals: Fast Scalper
Trading signals: Fast Scalper
  • 2013.10.08
  • Christian Beckmann
  • www.mql5.com
Fast Scalper Trading Signal for MetaTrader 5: social trading, mirror trading, copy trading and account monitoring
 

I think that the expert advisors is quite reliable if your logic depends on prices of closed bars.

If you need ticks to determine the action you have find your own solution ;-)

( I am still working on mine, collecting all ticks and then feed them into my price logic in sync with the timing of the strategy tester. Anyway one problem can't be solved that the order execution price is determined by the logic from the save terminal client.)

But I think an EA should handle all "suprises" in price moves in one minute bar anyway otherwise it would fail sooner or later.

Uwe

Documentation on MQL5: Standard Constants, Enumerations and Structures / Trade Constants / Order Properties
Documentation on MQL5: Standard Constants, Enumerations and Structures / Trade Constants / Order Properties
  • www.mql5.com
Standard Constants, Enumerations and Structures / Trade Constants / Order Properties - Documentation on MQL5
 
b3ck1:

Hi,

i spend a lot of time searching and testing ea's, i even started to learn mql5 to develop my own ea's.

But what i learned by now is that there is no reliable testing method for expert advisors - strategy tester does not behave like a real account and even the demo accounts i tried so far are not really reliable.

Example?

Look at  https://www.mql5.com/en/signals/15563 which was a test of a expert advisor i have developed. More than 50% growth within one and a half hour - not bad, right? 

Well, this ea is a kind of fake - it only works on accounts which are NOT using market execution and on a real account it does not trade at all because of requotes.

 

So, how can i test if a expert advisor will be profitable if even demo accounts are not reliable? 

mostly it depends on your strategy. mine came out almost the same as backtested and it is not a scalping system.

if you are scalping fast and taking little profits, requotes will kill your EA. Its too sensitive already. 

 
b3ck1:
 

So, how can i test if a expert advisor will be profitable if even demo accounts are not reliable? 

Why don’t you just open a live account especially for testing EA’s? If you have an account of 200$ for example you can set your EA that it will only trade if “free balance is >160$”. If as you say, you want to trade a small profit scalping EA and you set it to trade 0.01 Lot, this way you can lose max about 400pips which will be probably enough space for testing such a strategy.

 

I opened a real account to test this ea - but there were no trades made because the demo and the real account (even on the same broker) do not behave the same. 

the question was if real accounts are the only way to reliably testing an ea? for this you just have to buy the ea to afterwards see that it does not work... 

Documentation on MQL5: Standard Constants, Enumerations and Structures / Environment State / Account Properties
Documentation on MQL5: Standard Constants, Enumerations and Structures / Environment State / Account Properties
  • www.mql5.com
Standard Constants, Enumerations and Structures / Environment State / Account Properties - Documentation on MQL5
 
b3ck1:

the question was if real accounts are the only way to reliably testing an ea? for this you just have to buy the ea to afterwards see that it does not work... 

As far as I know the only proof if an EA will really work on a real account is to test it on a real account.

Yes, if you buy a EA, who was only tested on backtest or demo account you can never know if he's making as good results on a real account.

Some strategies perform very similar on demo and real accounts, others perform very different.
Documentation on MQL5: Standard Constants, Enumerations and Structures / Environment State / Account Properties
Documentation on MQL5: Standard Constants, Enumerations and Structures / Environment State / Account Properties
  • www.mql5.com
Standard Constants, Enumerations and Structures / Environment State / Account Properties - Documentation on MQL5
 
b3ck1:

I opened a real account to test this ea - but there were no trades made because the demo and the real account (even on the same broker) do not behave the same. 

the question was if real accounts are the only way to reliably testing an ea? for this you just have to buy the ea to afterwards see that it does not work... 

I guess the reason is because it needs some conditions that are not offered by many brokers in Real conditions. To my own understanding are the conditions needed of which mos of them can be found with DMA brokers who have minimal slippage.

No Commission
Tight Spreads from 0 to 1 pips on major pairs
Available in 5-digits pricing
Lightning fast sub-second execution
No re-quotes, No Dealing Desk
No trading restriction
Zero stop level - you can put TP and SL anywhere

Kind Regards