Backtesting/Optimization - page 88

New comment
 

big difference between MT4 optimization and backtesting results

Hello,

may you help me with understanding one thing connected with optimization and backtesting in MT4? In summary: results of backtest and optimization are much different.

I have done optimization and backtesting on the same VPS with the same instance of MT4 on my real account with the same broker.

The optimization was performed for some range of three values (FVB 3-7; SVB 50-72; VF 1-6). All the other things were the same - period (it was one full year), initial deposit, timeframe M15, currency pair EURUSD, way of optimizing (without genetic algorithm, with 'every tick' model) and lot size - they were the same for optimization and two backtests.

Based on optimization I have found these two values:

165 3405.39 80 2.29 42.57 1160.97 22.53% FVB=7 SVB=70 VF=3 LotSize=0.3

83 -423.42 547 0.98 -0.77 3460.25 52.65% FVB=5 SVB=60 VF=2 LotSize=0.3

The first one is the best result, the second one (in fact it was much lower in the list of results) was default settings of EA.

There were same warnings in the journal. There were five warnings of the first type (so not so many of those):

2013.02.18 19:33:26 TestGenerator: unmatched data error (low value 1.33561 at 2013.02.15 21:45 is not reached from the least timeframe, low price 1.33584 mismatches)

And really many exactly the same warnings connected with one single date 2012.06.22:

2013.02.18 19:33:07 TestGenerator: unmatched data error (volume limit 457 at 2012.06.22 20:45 exceeded)

So I guess these warning should not influence backtests and optimization.

I have performed backtest during weekend and optimization on Monday so some things could be different. However, this system is not the scalper so it should not influence these results much.

Results of backtesting were as follows:

FVB=7 SVB=70 VF=3 total net profit: 4307.70

FVB=5 SVB=60 VF=2 total net profit: 4976.00

As you can see, in first case (the best settings) profit was higher than in optimization by 26%.

In the second case (which was far away from optimal on the backtest - it indicated loss) there was higher profit than in the case of backtest of the best settings! What could be the reason?

Based on forwardtest I can say it was not loss.

I may also upload printscreens from optimization (all the plots resulting from optimization, table with optimization results) and from backtests (report, plots, table with trades).

I have also data from forwardtest so I can compare backtest on the same period and with the same settings as forwardtest.

Thanks in advance for guidelines!

 
akhmadfx:
This day i have created an EA on every new bar model EA,

i was tested that EA on strategy tester open price model and the result is very good,

but i think in the real trading we facing on every tick movement on price, so i decide to test on every tick model, and the result is very very bad.

this is some kind of MT4 bug or not. i still do not understand, somebody please answer

Try making your EA work only on open price too and forward test it that way. Since in the "open price mode" only the the price at the open of the bar is used (the open) if tests were OK then your froward test results should resemble your back test results

The "each tick mode" on the other hand is simulating ticks and it is very, very far from what actually happened with ticks and spreads (bid and ask prices). They say it is the most precise method but my advice is that you take it with 90% doubt when it comes to being "most precise"

 

Result is different tick and open price ???

This day i have created an EA on every new bar model EA,

i was tested that EA on strategy tester open price model and the result is very good,

but i think in the real trading we facing on every tick movement on price, so i decide to test on every tick model, and the result is very very bad.

this is some kind of MT4 bug or not. i still do not understand, somebody please answer

 
mladen:
Try making your EA work only on open price too and forward test it that way. Since in the "open price mode" only the the price at the open of the bar is used (the open) if tests were OK then your froward test results should resemble your back test results The "each tick mode" on the other hand is simulating ticks and it is very, very far from what actually happened with ticks and spreads (bid and ask prices). They say it is the most precise method but my advice is that you take it with 90% doubt when it comes to being "most precise"

Ok, thanks for your explanation Mladen,

On every new bar mode, trailing stop is active when every new bar formed,

but when i test on every tick , trailing stop is gonna wrong becouse my EA is ON EVRY NEW BAR mode.

kindly check my attachment above for detail, thanks

 

Please any share experience for the result of backtest with M1 99.9% model quality is the same or nearly the saame as real account?

 

Sorry,

which in the best broker for Back testing with MT4 for you?

Thanks

Stefano

 
stelore:
Sorry,

which in the best broker for Back testing with MT4 for you?

Thanks

Stefano

There is no such thing as best broker for back testing. The best thing you can do is to back test on your broker data and that way you will have a broader picture how will some EA perform on your broker (since all brokers have different data - you will not find one single broker that has exactly the same data as some other broker)

 

Hello to everyone.

I'm still relatively new in forex trading. I read about different EA. I learned about different EA logic. So I discovered that back tests are working differently for different types of logic. For Some EA they actually give you the idea of how they're going to work on your live account, and for another they might just help you to understand the algorithm and figure out the best settings. Well, that is how see the situation.

 

Backtest actually

Hello,

I know that there is so much threats talking about this, but I didn't find a good one where I can find it clarify and simple the solution.

I try following this steps: Forex Tick Data | Birt's EA review but I can't get the history to use it.

I can download the data from dukascopy (in CSV) and I use the Expert Advisor to convert it to the right format (Metatrader format).

After that, if I go to see the data at history (Cntrl + F2) I can see the data, but not all of them (I have downloaded a yera in 1M).

For that reason, I don't know how to get a good tick data to get 99% of backtest or at least, more than normal.

Thank you in advance, and if you move my thread, please, let me know to know the answer.

Is there some app? Or some video that can I learn it quick and simple?

Thank you again and sorry for my english.

Hermo

 

Strategy Tester Help

Hi All,

I just ran a strategy tester on IBFX for 6 months of data (Jan - June 2010) and after 2 weeks of running on a powerful pc (OMG that MT4 strategy testing thing in slow), I got a report that NO trades were placed. It basically did nothing -

Secondly, it told me data quality was 90%.

I am not sure why that happened, but the same EA works in demo account.

Any advice? Secondly, is there a better platform, tool, process or something else for speeding up the strategy testing and improve overall quality?

1...818283848586878889909192939495
New comment