How much time you consider enough to test a strategy before operate in real account and conditions?

New comment
 
  • 9% (7)
  • 9% (7)
  • 34% (28)
  • 34% (28)
  • 2% (2)
  • 12% (10)
Total voters: 82
 

asking my question in a poll format?

https://www.mql5.com/en/forum/19282

:) 

 

3 months

 

6 months would make me confident in the results of the EA.

 
doshur:

asking my question in a poll format?

https://www.mql5.com/en/forum/19282

:) 

To be honest, your topic is new for me, anyway we are in the same page, thanks for sharing this.
 

Forum on trading, automated trading systems and testing trading strategies

Evaluation period of EA

doshur, 2014.02.01 17:41

What is the period you run either demo live or real live to evaluate your EA?

there are cycle where the EA performs and not perform. So what is your ideal period before you throw it into the dustbin? 


 
6 months +
 
6 MONTHS IS YET ENOUGH I THINK  , BUT NEVER ACCURATE OF COURSE AS THE MARKET IS UNPREDICTABLE
 
depends on the strategy. Fully automated strategies that work mechanics are unknown have to be tested well before using them on live accounts. Strategies that use known patterns and indicators may be tested less. Some strategies you can use right away. (like strategies that support your manual trading) This question is way too general. Depends on the strategy and idea behind it.
 
angreeee:
depends on the strategy. Fully automated strategies that work mechanics are unknown have to be tested well before using them on live accounts. Strategies that use known patterns and indicators may be tested less. Some strategies you can use right away. (like strategies that support your manual trading) This question is way too general. Depends on the strategy and idea behind it.

Thanks for sharing but how do you know that a known past pattern will repeat again?

Don't you have to test it in the same way as fully automated strategies?

 
figurelli:

Thanks for sharing but how do you know that a known past pattern will repeat again?

Don't you have to test it in the same way as fully automated strategies?

You dont know. Thats why forex is game of probabilities. If it was that simple everyone would be rich out of just buying 1 EA.

How do you know that strategy wont just stop beeing profitable? I created a variation of a strategy generating channels out similar to keltner channel mathematics (that i learned recently that its very similar). It was profitable on every pair on many timeframes till the end of 2012 when it stopped beeing profitable on almost all pairs but long timeframes. You never know. for example my grid strategy will work on gbp/jpy with default settings until the trend will reverse or the drawdawn is going to be greater than in last 18 months. What is more to say? If you test it for 6 more months (i tested it for 4 months on live account, but that does not mean it will be profitable in the future if any of the two things ive written about occur) it will not change the simple fact that - it will be profuitable if the 2 things dont occur and it will wipe your account with default settings if these things occur. Testing such thing 6 months on a demo is a waste of time when you understand how the strategy is working and what the rules are. You either use default settings, prepare for unexpected or dont use it at all.

Another example is when i use my grid strategy in half-manual mode. I combine my knowledge with mathematic patterns of the strategy. It acts as i set it to act, all is represented by graphics. So if it fails, it is my mistake, not strategy's fault. Why would i want to test it for 6 months then?

Another example. With the strategy similar to keltner channel you can harvest daily chart of GBPUSD around SMA 400-500 form 2009.06-current getting even up to 1000x+ equity multiply in backtests. But you cannot say if the oscilation will continue or not. Testing six more months is not going to change that. Either you think GBPUSD will remain in oscilation mode or you think there is going to be a breakout. Again 6 month test that you would apply to fully automated strategies with hidden mathematics like lets say "angry beaver", are a waste of time here. 

Again. All depends on strategy, on what you want to accomplish, and on many different things. 

My English skills are not good enough to even explain 10% of different possibilities that im thinking of why this whole question is wrong. It is like asking: what is the best food for humans? For different men different food will be the best. The same with testing time for EAs. All depends on many factors.

12
New comment
Reason: