Best third-party strategy tester for MetaTrader 4 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

MetaTrader 4 Strategy Tester tutor 18 replies

Delete Expert Advisors from Strategy Tester (Metatrader 4) 0 replies

Metatrader 4 Strategy Tester Rollover 0 replies

MetaTrader 5 trading strategy tester 2 replies

MetaTrader strategy tester 0 replies

Attachments: Best third-party strategy tester for MetaTrader 4
Exit Attachments

Best third-party strategy tester for MetaTrader 4

  • #1
  • First Post: Jan 24, 2015 3:04am Jan 24, 2015 3:04am
Is anyone able to give some advice on the available options for using a third party application to test/optimize an EA (mt4)?

Metatrader 4 is limited to single core 32 bit operation, which is rubbish for the optimization feature in the strategy tester.
Metatrader 5 runs multi core 64 bit, so it cranks through optimizations. I love mt5 but....
I find mt4 a lot easier to develop in than mt5 and I can get access to better spreads & execution on mt4.

I have done the obvious google search, but nothing really appealed (I get disinterested when a trading product uses $ signs on their website).

So is anyone able to offer any advice or experience they might have please?

Thanks.
What's it all about? It's all about money.
  • #2
  • Edited 11:43pm Jan 27, 2015 6:36pm | Edited 11:43pm
No one is interested in this topic? Really?
I see threads on all sorts of nonsense, but when it comes to robust testing solutions, no one is interested. I can see why 90+% of you fail.

Anyway, if someone one day stumbles on this...

http://www.onlinestrategytester.com/ - looks pretty cool that they have all the data already (but what is the quality of that tick data?), however I am not a fan of uploading my ea to the internet!

http://forexsb.com/wiki/fsbpro/start - Seems to be a bunch of stuff I dont really need, but at least it does optimizations. Doesnt mention speed etc. More investigation required.

http://www.strategyquant.com/eaanalyzer/ - This looks like junk. All it does is analyse your completed backtest file. This is hopeless.

http://www.forexcontrolcenter.com/ - Same as above, all it does is import your statements from Metatrader. More junk.

http://www.forexstrategytester.com/f...ting-software/ - I like the look of this one, will download the trial version.

http://newsite.asirikuy.com/ - This looks really good. However they charge a monthly fee which I am not a fan of (The subscription costs 292 USD for the first year and then 161 USD for each year after that). I will happily pay for quality, but I want the software on my machine.


On the whole, pretty dissapointing at first glance.
Not one website made mention about how the tick data is handled or the use of the cpu. I will keep searching but I am thinking it might be back to MT5!
What's it all about? It's all about money.
  • #3
  • Jan 27, 2015 11:33pm Jan 27, 2015 11:33pm
I'm no longer fan of backtest, but i do had backtest acitivy back then, even it consume my trading hours .
have you ever try with them, http://www.tickstory.com/ , one single pair tick data sometimes on gigabytes.
my only problem is, pc spec couldnt handle the data convert and often crash.
well, you can always try different setting with backtest data, but the real data is the current market. BT purpose are just to check wether our ea logic works according to trading plan, result always vary from the FT process. spend more with forward test, enchance the setting while running your ea with real market data. while with BT only, no matter what tick data you've got, you will never had slippage data added, none of spread improvement collected as tick data anyways.
  • #4
  • Jan 28, 2015 12:42am Jan 28, 2015 12:42am
Quoting fxindikator
I'm no longer fan of backtest, but i do had backtest acitivy back then, even it consume my trading hours . have you ever try with them, http://www.tickstory.com/ , one single pair tick data sometimes on gigabytes. my only problem is, pc spec couldnt handle the data convert and often crash. well, you can always try different setting with backtest data, but the real data is the current market. BT purpose are just to check wether our ea logic works according to trading plan, result always vary from the FT process. spend more with forward...
Your EA should account for slippage and reject trades with excessive slippage so you cannot blame the tick data download for the problem.

Regarding Tickstory, I have never had any problem with data conversion.

Having tried most of the programs mentioned in Post # 2, my thoughts are that MT4 itself is fine for back testing. Just don't use MT4 history downloader. If you just want 90% back testing reliability, use SQ Tick Data Downloader which seems to be faster than Tickstory. Use "Configure" to export the tick data to CSV and tick the boxes to create CSV time series data and then import that into MT4. If you want 99.9% accuracy you will have to use Tickstory to do the tick import into MT4 and be sure to run your MT4 from Tickdata. You can reference the SQ tick data from Tickstory. Just rename the files to the Dukascopy format. To find out the format, just do a small download using Tickstory and look at the file name. It is Dukascopy format. Unfortunately SQ tick data filename has it's own format.

To summarize, MT4 itself is best since it is free and reliable. What is not reliable is the data being provided by MT4 history download. Therefore download your own data using a third party program such as SQ Tick Data Downloader or Tickstory.
  • #5
  • Jan 28, 2015 1:55am Jan 28, 2015 1:55am
Hi, thanks for your reply.
Whilst Tickstory data is the best that is easily accessible for the general public and I am thankful for, in real terms its actually very average.

If you open up the csv file and have a look at the data there are only handful of ticks for each interval. A sprinkle of the "real" volume. Im guessing its driven out of practicality as MT4 can only handle a 4 gig file max.

However, I think robust strategy testing is essential for any ea development. If you run an EA and it can't handle a 10+ year back test then you are just fooling yourself into thinking the strategy is robust. If your EA only survives 12 months before it crashes, then I hope you get lucky and withdraw enough money before you lose it all.

Running the ea on live data is essential but this is not where my focus is (slippage is inconsequential). I want to optimise a lot of data to look for convergences.
What's it all about? It's all about money.
  • #6
  • Jan 28, 2015 2:48am Jan 28, 2015 2:48am
Quoting Dave_C
{quote} To summarize, MT4 itself is best since it is free and reliable. What is not reliable is the data being provided by MT4 history download. Therefore download your own data using a third party program such as SQ Tick Data Downloader or Tickstory.
Dave_C

Thanks for your feedback on the links. It is quite disappointing to hear that MT4 is setting the bar. Such a shame as MT4 does have so much going for it, but fails miserably in this area.

Unless anyone has any other solutions looks like it will be mt5...
What's it all about? It's all about money.
  • #7
  • Feb 3, 2015 2:27am Feb 3, 2015 2:27am
Quoting Innate
No one is interested in this topic? Really? I see threads on all sorts of nonsense, but when it comes to robust testing solutions, no one is interested.
It appears that only few traders really care about reliable backtesting. I think the problem is that the backtesting process is hard to be done properly and takes time to study and master. It is difficult to program EA correctly. The MT backtester is slow and the backtest result for a expert is not good in most of the cases. It takes really good amount of time to make a logically correct expert with a good backtest performance and stats. That's why most of the traders gave up in the process and start gambling. Yes, yes. Trading without a correct backtest is GAMBLING.

Unfortunately there is no an easy and flawless solution. I'm forex professional and fighting with that problem for the last ten years. It appeared that the only way to me was to develop my own backtetsing engine.

I'm currently making a video series about creating Expert Advisors. You can find it in that channel: https://www.youtube.com/playlist?lis...WldbCHs10pT3qP

The series are going to cover the whole journey from establishing and testing a trading environment, making proper settings, gathering data, backtesting and analyzing forex strategies, preparing Expert Advisors and finally starting a real trade.

I may say I have a good experience in the field of backtesting and can discuss all topics about the process.

Quote
Just don't use MT4 history downloader. If you just want 90% back testing reliability, use SQ Tick Data Downloader which seems to be faster than Tickstory.

Dear Dave_C, when importing Tick Data to MetaTrader you really increase the "Modelling quality" number but this is the only goal you achieve. It doesn't improve you chance to make money on the live trade. Do you know why? The reason is very simple - the tick data downloaded with SQ Tick Data Downloader or Tickstory are comming from DuqasCopy, but they do not offer MT trading. The data dynamic is different, you cannot think the backest performed on data shaped by Ducas is reliable for your broker.
  • #8
  • May 30, 2015 11:59am May 30, 2015 11:59am
I really believe that spending to much time on backtesting is wasting time, backtests simply stated if a strategy has to be abandoned;
if backtest is positive, then we can go ahead in the development ---> WFA (IS/OOF) ---> Montecarlo ----> livetest with 0,01 lots
then increasing slowly lot sizing -----> adopt money management.

Hope it helps a bit
Xam
  • #9
  • Edited 3:33pm Aug 2, 2015 12:54am | Edited 3:33pm
Quoting Innate
http://www.forexstrategytester.com/f...ting-software/ - I like the look of this one, will download the trial version. ..Not one website made mention about how the tick data is handled or the use of the cpu.
Hi Innate, did you try this tool you liked (smart Forex tester)? It does use tick-by-tick data. The quotes you need to download from TrueFX.com - monthly .csv files. The only problem is that the files are really big for MS Excel to open them in full - if you want to view or edit the data, you need to use other software. You can also download a small utility from there to cut a month-long file into smaller chunks if need be. Personally, I don't backtest on longer than a week data.

BTW, in my opinion, trueFX looks really cool. Tradable quotes data for free. Archives for 15 main pairs since 2009 and also streaming quotes available via API.
Would be good to hear if anybody can comment on their data quality.
  • #10
  • Aug 2, 2015 11:30pm Aug 2, 2015 11:30pm
Quoting alku2011
{quote} Hi Innate, did you try this tool you liked (smart Forex tester)? It does use tick-by-tick data. The quotes you need to download from TrueFX.com - monthly .csv files. The only problem is that the files are really big for MS Excel to open them in full - if you want to view or edit the data, you need to use other software. You can also download a small utility from there to cut a month-long file into smaller chunks if need be. Personally, I don't backtest on longer than a week data. BTW, in my opinion, trueFX looks really cool. Tradable quotes...
Hi, sorry no I didnt investigate that much further than before.
I decided the only way forward (for me) was to step up to mt5 and use the strategy tester in that. Accordingly I havent looked back at mt4 and bothered with the hassle of downloading tick data etc etc...

I still fully support the notion that correctly performed backtesting is essential. If nothing else it cuts down development time of ea's.

Wish I could be of more help.
Best of luck!!
What's it all about? It's all about money.
  • #11
  • Edited 3:42am Aug 3, 2015 3:19am | Edited 3:42am
Quote
I decided the only way forward (for me) was to step up to mt5 and use the strategy tester in that

Are you testing on real tick data? I remember MT4 takes M1 bars and interpolates ticks from that, which makes the data artificial - too "smooth".

Thanks!
  • #12
  • Aug 3, 2015 8:56pm Aug 3, 2015 8:56pm
Quoting alku2011
{quote} Are you testing on real tick data? I remember MT4 takes M1 bars and interpolates ticks from that, which makes the data artificial - too "smooth". Thanks!
Hey alku,

No I am not using real tick data. In mt5 the tester uses the ohlc points and recreates the history. I still use the Every Tick mode to increase the accuracy. I have cross checked this before with 'real' tick data and the results were very very similar. Enough for me to not worry about chasing the elusive 'real' tick data. IMHO it was taking up too much of my time acquiring and loading the tick data instead of just testing and developing. My strategy is not heavily reliant on minute detail and mt5's tester gives me enough confidence that I am developing my strategy in the right direction.

Be wary about the quality of tick data out there. Unless you are harvesting it yourself it will quite likely not be the full picture. For example, if you open the csv file created from TickStory there is only a handful of ticks for each minute. This is not the real case. Instead there would hundreds or thousands of ticks every minute, depending on time and financial instrument.

mt4 vs mt5 coding is very similar nowadays. I am pleased I made the switch.
What's it all about? It's all about money.
  • #13
  • Aug 20, 2015 5:15am Aug 20, 2015 5:15am
I've been having problem with my backtesting until i solved the problem some ways..
Now, i backtest on Every ticks 70% faster like i do with Open Prices..
What more do i need?
I have Tick data from different sources..
Tickstory and StrategyQuant..
But, after seeing this truefx i will also give it a go, to make sure my strategy works best in all environment before i publish it to he public..
Nice Thread Here...
  • #14
  • Aug 20, 2015 8:53pm Aug 20, 2015 8:53pm
Quoting noobshow
I've been having problem with my backtesting until i solved the problem some ways.. Now, i backtest on Every ticks 70% faster like i do with Open Prices.. What more do i need? I have Tick data from different sources.. Tickstory and StrategyQuant.. But, after seeing this truefx i will also give it a go, to make sure my strategy works best in all environment before i publish it to he public.. Nice Thread Here...
Hey there, pleased to here you are having better success with your testing!

Can you please tell us what it is you did to make the back tester run 70% quicker? Thanks.
What's it all about? It's all about money.
  • #15
  • Aug 21, 2015 1:24am Aug 21, 2015 1:24am
Quoting Innate
{quote} Hey there, pleased to here you are having better success with your testing! Can you please tell us what it is you did to make the back tester run 70% quicker? Thanks.
I make sure all of my Codes have these lines
before the OrderSend() Line, so to delete all OrderSend Arrow and make backtesting faster.
Inserted Code
 if(IsTesting()) ObjectsDeleteAll();
If not debugging or verifying some codes, i use this also in my codes:
Inserted Code
if(!IsTesting) Print("This Is Needed While Demoing or Trading Live,Not needed here");
And, In All Of My Objects Drawing also, i do this:
Inserted Code
if(!IsTesting) ObjectCreate("Trading_Alone",OBJPROP_HLINE,0,Time[0],Ask+50*Point);
In Comments also, i do same
Inserted Code
if(!IsTesting)
Inserted Code
if(!IsTesting) Comment("Needed In Demo or Live Trading, Not In Backtest");
So, with all of this i get fast and faster backtest using Every Tick 70-80% Faster Than Before...

I think peoples need either an article on this or a Tutorial about it....
The ObjectsDeleteAll(); do the biggest function in all of my codes..
  • #16
  • Aug 21, 2015 1:31am Aug 21, 2015 1:31am
Quoting Innate
Is anyone able to give some advice on the available options for using a third party application to test/optimize an EA (mt4)? Metatrader 4 is limited to single core 32 bit operation, which is rubbish for the optimization feature in the strategy tester. Metatrader 5 runs multi core 64 bit, so it cranks through optimizations. I love mt5 but.... I find mt4 a lot easier to develop in than mt5 and I can get access to better spreads & execution on mt4. I have done the obvious google search, but nothing really appealed (I get disinterested when a trading...
Look into the cTrader platform cAlgo, they have a pretty good testing area. That tops MT backtester for sure, and is free with Pepperstone account.
All posts are my personal opinion
  • #17
  • Aug 27, 2015 7:13am Aug 27, 2015 7:13am
Quoting noobshow
{quote} I make sure all of my Codes have these lines before the OrderSend() Line, so to delete all OrderSend Arrow and make backtesting faster. if(IsTesting()) ObjectsDeleteAll(); If not debugging or verifying some codes, i use this also in my codes: if(!IsTesting) Print("This Is Needed While Demoing or Trading Live,Not needed here"); And, In All Of My Objects Drawing also, i do this: if(!IsTesting) ObjectCreate("Trading_Alone",OBJPROP_HLINE,0,Time[0],Ask+50*Point); In Comments also, i do same if(!IsTesting) if(!IsTesting) Comment("Needed In Demo...
This is really helpful, thank you!
What's it all about? It's all about money.
  • #18
  • Sep 6, 2015 5:23pm Sep 6, 2015 5:23pm
Quoting Innate
{quote} This is really helpful, thank you!
You're welcome...
  • #19
  • Nov 23, 2015 11:26am Nov 23, 2015 11:26am
So,

I'm use other way for speed testing.

Inserted Code
#define  _SHOW_     1
...
        #ifdef _SHOW_
        ObjectSet("UPLINE",OBJPROP_PRICE1,LastTop);
        ObjectSet("MIDDLEL",OBJPROP_PRICE1,LastBrick);
        ObjectSet("DNLINE",OBJPROP_PRICE1,LastBottom);
        #endif

if remove line with #define _SHOW_ , all grafic, alarms, etc.. is hide during compilation.

cheers
Milan
  • #20
  • Jun 3, 2016 12:02pm Jun 3, 2016 12:02pm
Hi Guys,
I have deeply tested one of the tool above mentioned (StreategyQuant). I have to say that I was initially a bit skeptical about the idea;
in a nutshell:

An algortim that can put together trading concepts+trading rules and generate with "Random way" or "Genetic algo" strategies that can be
analyzed, checked and traded with profit. Come on guy it can not work !

Then I strat working on it. I start reading, reading, reading. I refined my learing with best in class book, specificly in statistic, probability analysis.
I also start studing math concerning modelling simulation (I was already quite skilled in math modelling with stat tools but not specificly in trading strat).
After around a couple of months after having read all the doc concerninng Stragegyquant I decided to download the tools (stragyquant) and apply what I knew and what I learned about strategy generation; how to avoid overfitting, how to minimize optimization, how to skip optimization in order to gain in robustness. How to treat stat figures without compromise data sets , advanced techniques to apply T-Test to be consistent in the making decision process, etc....
......then guys, I have to admit the path was (and it is) not so easy, and require full dedication and lot of spare time, mainly during the night or early in the morning. So I download the tool and I start apply the option: "Startegy Generation", the tool generated lot of strats and I could start with a potential portfolio of around 100 strat potentially good (I didn't consider in that stage the profit as main driver but I analyzed the "text pseudocode" in order to check which kind of approach the strategy apply). So first step I did was to generate strategies and select based on specific creteria I think are important (I'm not avaialble here to share these criteria becasue I think are the core of my process and I want to work more on that). Than I decided to select only the strategy that were in my opinion logic and justified by the pattners behind the indicator used. I have to say theat the random strategy generated have all a good sense but in my opinion these strat tend to be overfitted due to the way in which are generated so have to be treated as they are (Potentially overfitted).
As a conseguenze of that I apply all the theory I learn to avoid overfitting and selecting stragegies and I did a cherry picking of strategy and I did a WFA (walk forward analysis) on new historical data in order to test how the model works.
As main parameeter I decided to use , in order to avoid overfitting and evaluate perf I choose the SharpeRatio (is not for sure my invention if you check in Financial literature, it is considered one good parameeter for that aspect even if likely the Sortino Ratio is a bit better... .. for my porpose I'm not looking for perfect sientific approach rather than an empiric running profitable approach and repeateble.
Finally I came out from the analysis with 10 strategy that I decided to test live-Demo with around 2-3 variation per strategy (the start are not optimized by me, so for each strat I decided to select a range of parameeter that I beleve should work - in that way I'm not doing overfitting but diversification among the same instrument). I eliminate too similar stretegy and decided for a last test before inplementing in the platform.
Last test: I did a tenetative to test the strategy found on different market - ETF, 10 Y Bond, Commodity, S&P. At the end few of the strategy found survive at the test. These 10 strat found (out of more that 10 millions, yes guys 10 millions strat explored with the Strategyquant tool) has been installed on a trading platform (MT4) and with the variation (in terms of parameeter - NOT OPTIMIZED REMEMBER, and in terms of market) I have now around 30 strat. My process require that each month I review performance to decide if a strat has to be revised or cencelled from portfolio. I will apply that model for one year and after that I will decide how to proceed. That is my experiance with that tool.
I think in that tool there is a genuine value that must be captured and used with a good knowled and preparation. Is not a tool that produce winning strategy; IT IS IN MY OPINION A TOOL THAT PRODUCE SMART PROCESSES TO BECAME PROFITABLE and require to study (not needed programming exp).
If you want you can test the portfolio implemented, In my case I did a scenario for my portfolio but and the outcame was nice.
In particular what I really like is that it pass the WFA (also here please be carful not to “overoptimized” the model). If you decide to define the best horizon to perform WFA, please select “good sense” criteria not the best WFA interval optimized (this is in my opionion overoptimization. At the end my experience , with any tool as for the manual strat I develop, tells me that less optimization is better, more robustness is better. We can make diversification also across parameters if we don’t overoptimize the parameters in themselves.
(as conclusion StrategyQuant is a pice of smart software that cost less than other very expensive tools; be experience in using it ! and you will have lot of potentialities more in your trading. In addition very easy to develope your idea in the strategy manual geberation)
Hope my experience is useful for you, if you need more detail on my process please contact me.
Cheers
Xam
0 traders viewing now