mt4 vs mt5 strategy tester | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Custom build PC for MT5 strategy tester's optimization? 3 replies

strategy builder/tester for new scalping strategy 1 reply

2009.03.11 15:41:15Tester memory handler: tester stopped because not enough memory 19 replies

MT4 Strategy Tester 0 replies

mt4 strategy tester 0 replies

Attachments: mt4 vs mt5 strategy tester
Exit Attachments

mt4 vs mt5 strategy tester

  • #1
  • First Post: Oct 30, 2013 5:17am Oct 30, 2013 5:17am
Hi,
I've tested macd sample ea with the same settings both on mt5 alpari tick data and activtrades mt4 tick data downloaded from dukascopy and imported with Birt's Tick data suite. Period is from 2012 10 20 to 2013 10 19.
Results are very different: on mt4 I have -637.78 profit. On mt5 I have -134.83?
Why this difference?
Are tick data on mt5 affordable? I would like to migrate to mt5 from mt4 due to the very time consuming and headache generating nature of mt4 strategy tester.
  • #2
  • Oct 30, 2013 5:24am Oct 30, 2013 5:24am
Quoting eugenioca
Hi, I've tested macd sample ea with the same settings both on mt5 alpari tick data and activtrades mt4 tick data downloaded from dukascopy and imported with Birt's Tick data suite. Period is from 2012 10 20 to 2013 10 19. Results are very different: on mt4 I have -637.78 profit. On mt5 I have -134.83? Why this difference? Are tick data on mt5 affordable? I would like to migrate to mt5 from mt4 due to the very time consuming and headache generating nature of mt4 strategy tester.
There is no way to test on MT5 with real tick data . . . the data you are getting in the MT5 Strategy Tester are simulated tick data made from the M1 data. Also the spread in MT5 is variable and taken from the M1 data so it can change from minute to minute, the Spread on MT4 ST is fixed for the duration of the test . . . unless you are using real Spread with the TDS ?
20 pips a day isn't too much to ask . . .
  • #3
  • Oct 30, 2013 6:12am Oct 30, 2013 6:12am
tnx for Your interesting answer, and yes I'm using real spread with TDS. I have the following problem. I've bought Birt's TDS and WF analyzer. I've payed a freelancer to code for me in mt4 2 eas based on 2 promising (imho) strategies. But before I go live I want to heavy backtest and wf test them with tickdata. I want to do this at least for the last 14 months tick data. If I want to perform such backtesting and wf tick data testing I need circa 4 hours for every parameter set I choose. So I need several days and I get lost in the process. (I have a centrino duo 2ghz pc). On the other hand I have seen that the mt5 strategy tester is hands off better than mt4 tester (talking about tester, not data). The first one uses cloud computing and anyway uses all the cores to backtest the eas. Now I don't know if it's better paying a coder to translate the 2 eas in mt5 or buying a faster pc, given that it's simply impossible to go further with backtesting and wf testing with my centrino pc and mt4 platform.
What's Your opinion about?
tnx
  • #4
  • Oct 30, 2013 7:40am Oct 30, 2013 7:40am
Quoting eugenioca
. What's Your opinion about? tnx
Your PC is too slow . . . can't you borrow a reasonable spec Laptop from someone to at least test the improvement you will get ? or perhaps rent one for a few days . . . or perhaps you could get a VPS and use that ? it is also possible that your EA is coded very badly, I made some simple changes to an EA for another User and improved the performance dramatically, 90 x faster: http://forum.mql4.com/54831
20 pips a day isn't too much to ask . . .
  • #5
  • Oct 30, 2013 8:11am Oct 30, 2013 8:11am
I have made some optimizing session even with an i3 at 2.1ghz. 150 optimization cycles (using open bar only) in 3h vs 4h 45m on centrino.... not a heck of improvement. However here is one of the eas. Feel free to check it if U want. The most important thing I want to know is: You confirm that using mt5 is a waste of time? That would be a pity cause I've discovered mt5 has a well projected strategy tester that make You able to backtest and wf test an ea all in one. It gives U the possibility to insert random delay in testing. Further the mt5 wizard makes You able to insert in your ea several trailing and money management strategies. But.... if the data it uses is rubbish... all these features are pointless. Isn't there a way to use the mt5 strategy tester and mql5 wizard architectures on truly tick data from dukascopy from instance?
Attached File(s)
File Type: mq4 media15_02_mod v1.1.mq4   45 KB | 436 downloads
  • #6
  • Oct 30, 2013 12:30pm Oct 30, 2013 12:30pm
Quoting eugenioca
Isn't there a way to use the mt5 strategy tester and mql5 wizard architectures on truly tick data from dukascopy from instance? {file}
Why do I need to repeat what I already said, just scroll up the page and read it again . . . http://www.forexfactory.com/showthre...44#post7058744


Oh dear . . . EA Builder produces bad code . . .

Why do you need to call this Indicator this many times for each tick ? did you read the thread I gave a link to ?

Inserted Code
double Buy1_1 = iCustom(NULL, PERIOD_M15, "AllAverages_v2.1 cc", 0,0,3,0,17, 0, Current + 1);
double Buy1_2 = iCustom(NULL, PERIOD_M15, "AllAverages_v2.1 cc", 0,0,13,0,17, 0, Current + 1);
double Buy2_1 = iCustom(NULL, PERIOD_M15, "AllAverages_v2.1 cc", 0,0,13,0,17, 0, Current + 1);
double Buy2_2 = iCustom(NULL, PERIOD_M15, "AllAverages_v2.1 cc", 0,0,34,0,17, 0, Current + 1);
double Buy3_1 = iCustom(NULL, PERIOD_M15, "Damiani_volatmeter", 13,50,1.3,true, 2, Current + 1);
double Buy3_2 = iCustom(NULL, PERIOD_M15, "Damiani_volatmeter", 13,50,1.3,true, 0, Current + 1);
double Buy4_1 =  Var3;
double Buy4_2 =  Var4 ;
double Buy5_1 = iCustom(NULL, PERIOD_D1, "AllAverages_v2.1 cc", 0,0,3,0,17, 0, Current + 1);
double Buy5_2 = iCustom(NULL, PERIOD_D1, "AllAverages_v2.1 cc", 0,0,13,0,17, 0, Current + 1);
double Buy6_1 = iCustom(NULL, PERIOD_D1, "AllAverages_v2.1 cc", 0,0,13,0,17, 0, Current + 1);
double Buy6_2 = iCustom(NULL, PERIOD_D1, "AllAverages_v2.1 cc", 0,0,34,0,17, 0, Current + 1);

double Sell1_1 = iCustom(NULL, PERIOD_M15, "AllAverages_v2.1 cc", 0,0,3,0,17, 0, Current + 1);
double Sell1_2 = iCustom(NULL, PERIOD_M15, "AllAverages_v2.1 cc", 0,0,13,0,17, 0, Current + 1);
double Sell2_1 = iCustom(NULL, PERIOD_M15, "AllAverages_v2.1 cc", 0,0,13,0,17, 0, Current + 1);
double Sell2_2 = iCustom(NULL, PERIOD_M15, "AllAverages_v2.1 cc", 0,0,34,0,17, 0, Current + 1);
double Sell3_1 = iCustom(NULL, PERIOD_M15, "Damiani_volatmeter", 13,50,1.3,true, 2, Current + 1);
double Sell3_2 = iCustom(NULL, PERIOD_M15, "Damiani_volatmeter", 13,50,1.3,true, 0, Current + 1);
double Sell4_1 =  Var3 ;
double Sell4_2 =  Var4 ;
double Sell5_1 = iCustom(NULL, PERIOD_D1, "AllAverages_v2.1 cc", 0,0,3,0,17, 0, Current + 1);
double Sell5_2 = iCustom(NULL, PERIOD_D1, "AllAverages_v2.1 cc", 0,0,13,0,17, 0, Current + 1);
double Sell6_1 = iCustom(NULL, PERIOD_D1, "AllAverages_v2.1 cc", 0,0,13,0,17, 0, Current + 1);
double Sell6_2 = iCustom(NULL, PERIOD_D1, "AllAverages_v2.1 cc", 0,0,34,0,17, 0, Current + 1);
20 pips a day isn't too much to ask . . .
  • #7
  • Oct 30, 2013 1:22pm Oct 30, 2013 1:22pm
ok, I got Your 7v member valuable opinion. I'm only disappointed becouse I'm forced to use a platform (mt4) that allows me to load truly tick data (for example dukaskopy tick data) but it's without doubt primitive compared to other platforms last releases. About this part of my ea code, there is a reason for that. I use this site:
http://sufx.core.t3-ism.net/ExpertAdvisorBuilder/
in order to build new strategies and after that I insert this chunk of code in the body of an ea that allows me to open several positions and has more features than the eas generated with expert advisor builder. I already knew this part of the ea isn't properly coded.... but I'm not able to code mt4 and I can't hire a coder every time I want to try a new strategy...
  • #8
  • Oct 30, 2013 2:36pm Oct 30, 2013 2:36pm
Quoting eugenioca
ok, I got Your 7v member valuable opinion. I'm only disappointed becouse I'm forced to use a platform (mt4) that allows me to load truly tick data (for example dukaskopy tick data) but it's without doubt primitive compared to other platforms last releases. About this part of my ea code, there is a reason for that. I use this site: http://sufx.core.t3-ism.net/ExpertAdvisorBuilder/ in order to build new strategies and after that I insert this chunk of code in the body of an ea that allows me to open several positions and has more features...
hi eugenioca,

I do not know your strategy but looking at those indicator calls and timeframes & periods used I would ask if you REALLY need to even use tick data?
  • #9
  • Oct 30, 2013 3:55pm Oct 30, 2013 3:55pm
Quoting eugenioca
ok, I got Your 7v member valuable opinion. I'm only disappointed becouse I'm forced to use a platform (mt4) that allows me to load truly tick data (for example dukaskopy tick data) but it's without doubt primitive compared to other platforms last releases. About this part of my ea code, there is a reason for that. I use this site: http://sufx.core.t3-ism.net/ExpertAdvisorBuilder/ in order to build new strategies and after that I insert this chunk of code in the body of an ea that allows me to open several positions and has more features...
I was disappointed when I found that MT5 can't use real tick data . . . but it can't, you can't even import your own data even if it's M1 data . . . it doesn't use .hst files like MT4, it's data is stored encrypted.

You don't need to believe me, read for yourself: http://www.mql5.com/en/forum/11241 http://www.mql5.com/en/forum/11387

If you follow the link I gave ( http://forum.mql4.com/54831 ) you will see the code changes I made to improve the performance of an EA by a factor of 90 x simply by changing the iCustom() calls to once per bar from every tick . . . spend a little time, do a little research, modify a little code.
20 pips a day isn't too much to ask . . .
  • #10
  • Oct 31, 2013 2:40am Oct 31, 2013 2:40am
Quote
I do not know your strategy but looking at those indicator calls and timeframes & periods used I would ask if you REALLY need to even use tick data?
I don't know if I really need to use tick data too. What I know for sure is that testing an ea with data other than tick data make you believe in unicorns (or in gains in the order of 500% yearly). What I'm looking for is 1% monthly but rock solid and with a "surprise proof" drawdown rate. My ea trades in m5 tf but only in a market trending in longer tf. I prefer to look at tick data and walk forward testing becouse tick data and wf testing are less subject to hoodwink you.

Quote
You don't need to believe me, read for yourself: http://www.mql5.com/en/forum/11241 http://www.mql5.com/en/forum/11387

If you follow the link I gave ( http://forum.mql4.com/54831 ) you will see the code changes I made to improve the performance of an EA by a factor of 90 x simply by changing the iCustom() calls to once per bar from every tick . . . spend a little time, do a little research, modify a little code.

I believe You and I'm willing to learn coding but days have only 24hrs. However if You're on hire and are cheap You can pvt msg me... ;-)
  • #11
  • Oct 31, 2013 3:06am Oct 31, 2013 3:06am
Quoting eugenioca
{quote} I don't know if I really need to use tick data too. What I know for sure is that testing an ea with data other than tick data make you believe in unicorns (or in gains in the order of 500% yearly). What I'm looking for is 1% monthly but rock solid and with a \"surprise proof\" drawdown rate. My ea trades in m5 tf but only in a market trending in longer tf. I prefer to look at tick data and walk forward testing becouse tick data and wf testing are less subject to hoodwink you. {quote} I believe You and I'm willing to learn coding but...

I would say that your belief that tick data will add a robustness to your system that uses M5 timeframe is unfortunately the most resource expensive way to achieve the result. Tick data is still encompassed inside bar data so utilizing the O,H,L,C covers all ticks and and therefore provides a greater shock proofing than tick data if used correctly. Bar data is far quicker to process and whilst it covers all the price shocks it unfortunately just like tick data will not replicate liquidity and fill issues. Those that cry what about the spread I would suggest that if the random spread widening is a tipping point between success and failure of a system then the system is flawed anyway.

I only use tick data in systems for scalping where the TF becomes irrelevant, for all other systems I use bar processing.
  • #12
  • Oct 31, 2013 4:06am Oct 31, 2013 4:06am
oK,
very interesting and very useful to know it. So in strategy tester I would have to check opening price method, right?
  • #13
  • Last Post: Oct 31, 2013 4:16am Oct 31, 2013 4:16am
Quoting eugenioca
oK, very interesting and very useful to know it. So in strategy tester I would have to check opening price method, right?
depends how you have developed the EA, but yes using opening prices means your EA will be given a TICK only once per bar at the open and it's value will be the open price

but even if you want to use tick data, unless you need a mid-bar indicator value (most are useless mid-bar) then as Raptor also pointed out you should be checking for a new bar condition anyway and only updating the indicators once every bar relevant to the timeframe called in the indicator, ie: you should only need to call a daily timeframe indicator once a day.
0 traders viewing now