Discordance. I wish I could identify where's the fail from ? - page 2

To add comments, please log in or register
Icham Aidibe
11542
Icham Aidibe  
Marco vd Heijden:

Experience comes with practice, not with back testing... that can last forever and is completely rigged, and unnecessary.

Ma-the-ma-tics. Not philosophy, Marco. Let's take as an example, 2 MA, how do you choose the period while live trading ?

  • Feelings : adjusting day by day, each hour or 2H if in M1
  • Tester : best result, with best profit on 3 months
Stanislav Korotky
28727
Stanislav Korotky  
Icham Aidibe:
@Stan We won't troll there about what the strategy tester is useful & made for
I'm not sure I get you. If you mean that the suggestion about tick generation is not related to your case, please provide more info. Nobody can help without knowing some more details about your EA and environment.
Icham Aidibe
11542
Icham Aidibe  
Stanislav Korotky:
I'm not sure I get you. If you mean that the suggestion about tick generation is not related to your case, please provide more info. Nobody can help without knowing some more details about your EA and environment.
Well, uncertain it could be due to the tick generations process, nor related to the execution speed of mem vps/server. I was wondering if someone there was aware of any other factor ... that could make "the factory test" discordant with a try given in live, on the same period, same date, quite same conditions as it's in live.
Alain Verleyen
37719
Alain Verleyen  
Icham Aidibe:
Well, uncertain it could be due to the tick generations process, nor related to the execution speed of mem vps/server. I was wondering if someone there was aware of any other factor ... that could make "the factory test" discordant with a try given in live, on the same period, same date, quite same conditions as it's in live.

Of course the tick generation process is an important parameter, isn't obvious ?

Stanislav said you it's now possible to use real ticks, don't you think it could lead to better results that simulated ticks ?

Of course, that's not the only parameter. In backtest there is no slippage, requotes, no connection issue, no latency, the spread is fixed, and probably more. I am really wondering why you expect to get the same results.

Icham Aidibe
11542
Icham Aidibe  
Alain Verleyen:

Of course the tick generation process is an important parameter, isn't obvious ?

Stanislav said you it's now possible to use real ticks, don't you think it could lead to better results that simulated ticks ?

Of course, that's not the only parameter. In backtest there is no slippage, requotes, no connection issue, no latency, the spread is fixed, and probably more. I am really wondering why you expect to get the same results.

real ticks from the live chart, you won't set an ea just upon, i think we can fall agree upon that fact. as an amends, there's the "each tick mode" from the tester. i've reduced latency to 3ms and even allowed a 1seconds delay to cover that. so same same same results, not really, but alike, yes. spread is fixed on my test. i'm checking ... maybe the slippage ...

but it's not the slippage. i bet upon perturbated conditions. you ear what i'm saying l'ange, you're belgian. une affaire d'élevage, paraît il.

Alain Verleyen
37719
Alain Verleyen  
Stanislav Korotky:
Both MT4 and MT5 (latest builds) support testing on real ticks history (if your broker provides this data).
MT4 with real ticks ? you mean with a third party tool, right ?
Icham Aidibe
11542
Icham Aidibe  
Alain Verleyen:
MT4 with real ticks ? you mean with a third party tool, right ?
Actually I'm talking of MT5. I've try to run on MT4, but as soon as I've noticed that even with a third party tool it wasn't easy to get full data, I decided to continue working with MT5 - "at least it's the broker historical ..."
Stanislav Korotky
28727
Stanislav Korotky  
Alain Verleyen:
MT4 with real ticks ? you mean with a third party tool, right ?
AFAIK, third party tools can be used to make the task more user friendly, but they are not required in general. The essence of the method is to replace .fxt file in tester/history folder manually with a file with real tick data, which should be downloaded from broker or collected locally. The method was discussed on the forums many times.
Stanislav Korotky
28727
Stanislav Korotky  
Icham Aidibe:
Actually I'm talking of MT5. I've try to run on MT4, but as soon as I've noticed that even with a third party tool it wasn't easy to get full data, I decided to continue working with MT5 - "at least it's the broker historical ..."
For your problem, it makes sense to start with a minimal data range of testing (with not more than a couple of trades in it) and analyze differences between tester deals and live report.
12
To add comments, please log in or register