Backtesting opinion - page 2

New comment
 

I just remembered I have this thread https://www.mql5.com/en/forum/7841

There are a few 2012 participant which posted their testing report. At least all are profitable to be able to take part in the ATC2012.

Only a few survived the race. The optimization are too curve fitted?

 

newdigital:

good backtesting results are not guarantee the future profit because market is changed all the time. Besides, good backtesting results are not guaranted against forex scam.

Backtesting is just an instrument only for coders and traders. Just my opinion sorry.

Good Point.

Ubzen:

Given the information provided by You and doshur there's no way of telling which system would out-perform the other in the Future. Both traders must assume the risk of losing some amount of their deposit in the Future. I would personally trust a system which performed well over 10-years vs 5-years (only).

Again there's no additional details like #-of-Trades || Return on Investment || Drawdown .. etc. So the way I look at it is ... not only can it handle conditions within the last 5-years ... it can also handle conditions before than. I really cannot see how recent parameters trumps the all-time parameters unless it somehow have additional benefits.

Do  you agree on the point mention by newdigital above? 

ATC 2012 Participants
ATC 2012 Participants
  • www.mql5.com
The verification of your Expert Advisor is complete.
 
doshur:

I just remembered I have this thread https://www.mql5.com/en/forum/7841

There are a few 2012 participant which posted their testing report. At least all are profitable to be able to take part in the ATC2012.

Only a few survived the race. The optimization are too curve fitted?

 

Good Point.

Do  you agree on the point mention by newdigital above? 

Yeah... I can agree with that.
 
Ubzen:
Yeah... I can agree with that.

In that case, I believe the market changes over the 10 years.

So i can sort of assume the EA that survive 10 years of back testing to be robust and sort of be profitable in live?

 

But as others say, there is still no guarantee. Problems like re-quotes, delay made the EA fail?

 

I should optimize my EA every now and then to keep up with the market changes? 

 
doshur: In that case, I believe the market changes over the 10 years. So i can sort of assume the EA that survive 10 years of back testing to be robust and sort of be profitable in live? But as others say, there is still no guarantee. Problems like re-quotes, delay made the EA fail? I should optimize my EA every now and then to keep up with the market changes?
Those are the types of questions only you as the trader can answer. Like I've said before, I would use testing to compare 2-different systems and move forward with the one with the best results.... because thats the best I can do at that time. But anything can happen and no-one can say what will or should happen. 
 
doshur:


I should optimize my EA every now and then to keep up with the market changes? 

In my opinion . . .  curve fitting to past data does not guarantee that your curve will fit to future data.  If you can't predict price, so need to curve fit, what makes you think you can predict the curve you need to fit for the next 1/7/30 days ?
 
so how much value is there in back testing?
 
doshur: so how much value is there in back testing?

The same value there is in Live testing. Back-Testing just allows you to evaluate much of the conditions faster(not all conditions). Just because someone made money trading in the past doesn't mean they're going to continue making money in the future. If you have two_signals to evaluate ... Signal_A has high-returns with very low draw-down && Signal_B has low-returns with very high draw-down. Which Signal should you choose to go with ... A || B. Of course you go with Signal_A. But then the next week, Signal_A goes bankrupt. What do you say ... Live results are useless?

You make a decision with the information you have now ... not whats going to happen in the future. 

 
doshur:
so how much value is there in back testing?
That depends on what you are expecting to determine from your backtesting . . .  if you apply a scientific approach - experimentation,  or an engineering approach - testing,  you will understand the limits of what the Strategy Tester can tell you and you will be able to understand what it is telling you. 
 
My current standard for testing EA is it must be profitable in my back test. Then profitable in forward back test.

But my period for back rest is a mere 8 months.  Forward back test is also 8 months.

I believe that market changes over the years so I feel 10 years is too much. That's why is use a mere 8 months. 
 
doshur:
My current standard for testing EA is it must be profitable in my back test. Then profitable in forward back test.

But my period for back rest is a mere 8 months.  Forward back test is also 8 months.

I believe that market changes over the years so I feel 10 years is too much. That's why is use a mere 8 months. 
How do you know that your curve fitted for the last 8 months is going to "fit" the next month, 2 months, 6 months ?  or do you repeat this daily and have a moving daily fitted curve ?  how do you know it will fit for the next day ?
12345
New comment