You are not logged in or registered. Please login or register to use the full functionality of this board...
J Talon LLC Expert Advisor Solutions Forum
What hardware is adequate for tick backtesting? - Printable Version

+- J Talon LLC Expert Advisor Solutions Forum (https://www.jtatoday.info/forum)
+-- Forum: Price Action (https://www.jtatoday.info/forum/forumdisplay.php?fid=87)
+--- Forum: PA - General Discussion (https://www.jtatoday.info/forum/forumdisplay.php?fid=94)
+--- Thread: What hardware is adequate for tick backtesting? (/showthread.php?tid=20363)

Pages: 1 2


What hardware is adequate for tick backtesting? - llforex - 05-24-2015

Dear d4v3,

Now that I have switched to tick data from fxcm data for backtesting of PA, I realized the backtesting now is so much slower:  In 24 hours with fxcm I could backtest 15 years (2000-2015), with tick data I could hardly test several months, not even one year.  With this snail speed, my beard would be 9 inches long before I could finish the pairs with variations of the preset!

The solution for this problem is simultaneous backtesting of many pairs with different preset variations.  The time has come to upgrade my hardware.  But I know so little about computer hardware and what they can do.  On the one hand, there is no need to waste money, on the other hand I wish to get the testing done timely.  

Would you please advise:  for my purpose of simultaneously running 20 mt4 backtesting, at least running 10 mt4 backtesting, what minimum computer configuration is necessary?  Would the following configuration get the job done?

CUSTOM BUILT INTEL I7-4790 COMPUTER SYSTEM
  • Intel Core i7 Processor i7-4790 3.6GHz 8MB QUAD CORE

  • INTEL COPPER HEAVY DUTY LGA 1150 COOLING FAN

  • ZEROtherm Advanced ZT100 Thermal Grease

  • PROFESSIONAL WIRING

  • ASUS B85M-G LGA 1150 Intel B85 HDMI SATA 6Gb/s USB 3.0 uATX Intel Motherboard

  • 16GB (2 x 8GB) 240-Pin DDR3 1600MHz (PC3 12800) Dual You are not allowed to view links. Register or Login to view. (HIGH PERFORMANCE)

  • (2X) MEMORY HEAT SPREADERS

  • WD Green 3TB 3.5" Internal Hard Drive - OEM Packaging #WD30EZRX

  • 24X DUAL LAYER DVD-RW

  • INTEL HD 1GB HDMI/DVI 1080p PCI-EXPRESS VIDEO (ONBOARD)

  • REALTEK 8-You are not allowed to view links. Register or Login to view. DIGITAL SOUND ONBOARD

  • REALTEK 10/100/1000 Gigabit Network Card (onboard)

  • Antec NEW SOLUTION SERIES VSK-4000 Black SGCC steel ATX Mid Tower Computer Case

  • (2X) 120MM CASE FANS

  • hec X ORION 585 WATT POWER SUPPLY

  • 3-Year Limited Warranty Plan with Lifetime of free USA based Support Custom Hand Wiring For Ultimate Air Flow, Assembled in Cleveland, Ohio, USA

  • FED-EX GROUND SHIPPING (1-7 BUSINESS DAYS)
Thank you very much for your advice.

llforex


RE: What hardware is adequate for tick backtesting? - hopfi2k - 05-24-2015

(05-24-2015, 05:14 PM)llforex Wrote: You are not allowed to view links. Register or Login to view.Dear d4v3,

Now that I have switched to tick data from fxcm data for backtesting of PA, I realized the backtesting now is so much slower:  In 24 hours with fxcm I could backtest 15 years (2000-2015), with tick data I could hardly test several months, not even one year.  With this snail speed, my beard would be 9 inches long before I could finish the pairs with variations of the preset!

The solution for this problem is simultaneous backtesting of many pairs with different preset variations.  The time has come to upgrade my hardware.  But I know so little about computer hardware and what they can do.  On the one hand, there is no need to waste money, on the other hand I wish to get the testing done timely.  

Would you please advise:  for my purpose of simultaneously running 20 mt4 backtesting, at least running 10 mt4 backtesting, what minimum computer configuration is necessary?  Would the following configuration get the job done?

CUSTOM BUILT INTEL I7-4790 COMPUTER SYSTEM
  • Intel Core i7 Processor i7-4790 3.6GHz 8MB QUAD CORE

  • INTEL COPPER HEAVY DUTY LGA 1150 COOLING FAN

  • ZEROtherm Advanced ZT100 Thermal Grease

  • PROFESSIONAL WIRING

  • ASUS B85M-G LGA 1150 Intel B85 HDMI SATA 6Gb/s USB 3.0 uATX Intel Motherboard

  • 16GB (2 x 8GB) 240-Pin DDR3 1600MHz (PC3 12800) Dual You are not allowed to view links. Register or Login to view.[/url] (HIGH PERFORMANCE)

  • (2X) MEMORY HEAT SPREADERS

  • WD Green 3TB 3.5" Internal Hard Drive - OEM Packaging #WD30EZRX

  • 24X DUAL LAYER DVD-RW

  • INTEL HD 1GB HDMI/DVI 1080p PCI-EXPRESS VIDEO (ONBOARD)

  • REALTEK 8-You are not allowed to view links. Register or Login to view.[/url] DIGITAL SOUND ONBOARD

  • REALTEK 10/100/1000 Gigabit Network Card (onboard)

  • Antec NEW SOLUTION SERIES VSK-4000 Black SGCC steel ATX Mid Tower Computer Case

  • (2X) 120MM CASE FANS

  • hec X ORION 585 WATT POWER SUPPLY

  • 3-Year Limited Warranty Plan with Lifetime of free USA based Support Custom Hand Wiring For Ultimate Air Flow, Assembled in Cleveland, Ohio, USA

  • FED-EX GROUND SHIPPING (1-7 BUSINESS DAYS)
Thank you very much for your advice.

llforex

llforex,

First, it is important to understand, that backtesting is only limited in its reliability. IF you are backtesting, use tickdata (You are not allowed to view links. Register or Login to view. is a quite helpfull tool) or don't do backtesting at all! You can look at it as: a slow backtest with plenty data gives you a much better and more reliable picture.

Second, running like 9 or 10 backtest in parallel isn't that realistic too. From my knowledge, the Metatrader 4 platform doesn't make use of multiple cpu's/cores. Therefor if you really MUST do parallel backtesting, I'd suggest buying an eight-core CPU system, and run each Metatrader 4 instance in a virtual machine (e.g. VirtualBox). Anyhow, 8 core CPUs will be costy - like the system you mentioned.

Can you explain, why you need to run so many backtests parallel? Can't you run just on optimization test and give it some time?

Cheers,
Andy


RE: What hardware is adequate for tick backtesting? - llforex - 05-25-2015

(05-24-2015, 09:51 PM)hopfi2k Wrote: You are not allowed to view links. Register or Login to view.llforex,

First, it is important to understand, that backtesting is only limited in its reliability. IF you are backtesting, use tickdata (You are not allowed to view links. Register or Login to view. is a quite helpfull tool) or don't do backtesting at all! You can look at it as: a slow backtest with plenty data gives you a much better and more reliable picture.

Second, running like 9 or 10 backtest in parallel isn't that realistic too. From my knowledge, the Metatrader 4 platform doesn't make use of multiple cpu's/cores. Therefor if you really MUST do parallel backtesting, I'd suggest buying an eight-core CPU system, and run each Metatrader 4 instance in a virtual machine (e.g. VirtualBox). Anyhow, 8 core CPUs will be costy - like the system you mentioned.

Can you explain, why you need to run so many backtests parallel? Can't you run just on optimization test and give it some time?

Cheers,
Andy

Hi, Andy,

d4v3 said PA EA could possibly be used on other pairs too.  Tick data testing is very slow.  If I test more than one pair at the same time, I have hope to learn if the EA works on other pairs without waiting forever.  I don't know how to use optimization test.  Need to learn that.

Thanks.

llforex


RE: What hardware is adequate for tick backtesting? - Ratz - 05-25-2015

(05-24-2015, 09:51 PM)hopfi2k Wrote: You are not allowed to view links. Register or Login to view.llforex,

First, it is important to understand, that backtesting is only limited in its reliability. IF you are backtesting, use tickdata (You are not allowed to view links. Register or Login to view. is a quite helpfull tool) or don't do backtesting at all! You can look at it as: a slow backtest with plenty data gives you a much better and more reliable picture.

Second, running like 9 or 10 backtest in parallel isn't that realistic too. From my knowledge, the Metatrader 4 platform doesn't make use of multiple cpu's/cores. Therefor if you really MUST do parallel backtesting, I'd suggest buying an eight-core CPU system, and run each Metatrader 4 instance in a virtual machine (e.g. VirtualBox). Anyhow, 8 core CPUs will be costy - like the system you mentioned.

Can you explain, why you need to run so many backtests parallel? Can't you run just on optimization test and give it some time?

Cheers,
Andy


Some tick testing companies/programs allow you to open several platforms at once and have each back test a different setting (or full optimizations), this way you don't need several EURUSD fxt's to run them as they can all run off of just the one file. Sometimes I run up around 10 opts this way. What works good is when opting (which may take a week or 2 to complete) when you find a interesting setting, open another platform and run a smaller opt of that setting. Who knows it may be the best opt from the larger opt and you'll have a head start on it Wink  


RE: What hardware is adequate for tick backtesting? - llforex - 05-26-2015

(05-25-2015, 05:52 AM)Ratz Wrote: You are not allowed to view links. Register or Login to view.Some tick testing companies/programs allow you to open several platforms at once and have each back test a different setting (or full optimizations), this way you don't need several EURUSD fxt's to run them as they can all run off of just the one file. Sometimes I run up around 10 opts this way. What works good is when opting (which may take a week or 2 to complete) when you find a interesting setting, open another platform and run a smaller opt of that setting. Who knows it may be the best opt from the larger opt and you'll have a head start on it Wink  

Thanks, Ratz. The best proof is a live account, next is a demo.  The only reason I use backtesting is it is fast.  If a preset can't even pass backtesting, then it has no value.  If it does, it is worth setting up a demo account for the preset.  Sadly backtesting using tick data takes away the speed advantage.  I just try to keep the tick accuracy and solve the speed problem.

Would an entry level server allow me to open multiple mt4 platforms and backtesting many pairs at the same time?

Thanks.

llforex


RE: What hardware is adequate for tick backtesting? - Ratz - 05-26-2015

(05-26-2015, 01:53 AM)llforex Wrote: You are not allowed to view links. Register or Login to view.Thanks, Ratz. The best proof is a live account, next is a demo.  The only reason I use backtesting is it is fast.  If a preset can't even pass backtesting, then it has no value.  If it does, it is worth setting up a demo account for the preset.  Sadly backtesting using tick data takes away the speed advantage.  I just try to keep the tick accuracy and solve the speed problem.

Would an entry level server allow me to open multiple mt4 platforms and backtesting many pairs at the same time?

Thanks.

llforex

Sorry liforex, I have never done this way so can't comment on it ~
--can anyone answer?...


RE: What hardware is adequate for tick backtesting? - d4v3 - 05-26-2015

I do not use tick data for back testing so can't comment on the speeds that you're seeing, but I don't understand how it can be that much slower - tick data reads every individual tick from file whereas MQL data reads only M1 open, high, low, close and interpolates the individual ticks, but each processes roughly the same number of ticks during the back test, so I would not expect a great difference in the speed - unless there's something in the tick data set up which slows down the whole process.

As for backtesting on multiple MT4 applications I would say that you will be limited by the number of cores in the processor - on a quad core machine you could run 4 MT4 applications and each would use the full 25% of the processor allowed (assuming you had no other programs running) - adding another MT4 application would mean that 2 would have to share one core, so both would use 12.5%, meaning they would run half as fast.


RE: What hardware is adequate for tick backtesting? - llforex - 05-27-2015

(05-26-2015, 10:35 AM)d4v3 Wrote: You are not allowed to view links. Register or Login to view.I do not use tick data for back testing so can't comment on the speeds that you're seeing, but I don't understand how it can be that much slower - tick data reads every individual tick from file whereas MQL data reads only M1 open, high, low, close and interpolates the individual ticks, but each processes roughly the same number of ticks during the back test, so I would not expect a great difference in the speed - unless there's something in the tick data set up which slows down the whole process.

As for backtesting on multiple MT4 applications I would say that you will be limited by the number of cores in the processor - on a quad core machine you could run 4 MT4 applications and each would use the full 25% of the processor allowed (assuming you had no other programs running) - adding another MT4 application would mean that 2 would have to share one core, so both would use 12.5%, meaning they would run half as fast.

It would be wonderful if tick data backtesting were just as fast as, say, fxcm data backtesting.  But the snail speed is so obvious.  Further even between pairs, tick data backtesting speed can be different, i.e. NZDCAD is twice as fast as EURUSD.

Anyway, if I invest to upgrade my computer, how many simultaneous mt4 backestings can the following configuration comfortably handle?

CUSTOM BUILT INTEL I7-4790 COMPUTER SYSTEM

  • Intel Core i7 Processor i7-4790 3.6GHz 8MB QUAD CORE

  • INTEL COPPER HEAVY DUTY LGA 1150 COOLING FAN

  • ZEROtherm Advanced ZT100 Thermal Grease

  • PROFESSIONAL WIRING

  • ASUS B85M-G LGA 1150 Intel B85 HDMI SATA 6Gb/s USB 3.0 uATX Intel Motherboard

  • 16GB (2 x 8GB) 240-Pin DDR3 1600MHz (PC3 12800) Dual You are not allowed to view links. Register or Login to view.[/url] (HIGH PERFORMANCE)

  • (2X) MEMORY HEAT SPREADERS

  • WD Green 3TB 3.5" Internal Hard Drive - OEM Packaging #WD30EZRX

  • 24X DUAL LAYER DVD-RW

  • INTEL HD 1GB HDMI/DVI 1080p PCI-EXPRESS VIDEO (ONBOARD)

  • REALTEK 8-You are not allowed to view links. Register or Login to view.[/url] DIGITAL SOUND ONBOARD

  • REALTEK 10/100/1000 Gigabit Network Card (onboard)

  • Antec NEW SOLUTION SERIES VSK-4000 Black SGCC steel ATX Mid Tower Computer Case

  • (2X) 120MM CASE FANS

  • hec X ORION 585 WATT POWER SUPPLY

  • 3-Year Limited Warranty Plan with Lifetime of free USA based Support Custom Hand Wiring For Ultimate Air Flow, Assembled in Cleveland, Ohio, USA

  • FED-EX GROUND SHIPPING (1-7 BUSINESS DAYS)
Thank you very much for your advice.

llforex


RE: What hardware is adequate for tick backtesting? - d4v3 - 05-27-2015

As I said previously I would think that you will get the maximum speed by running one MT4 application per core - so a quad core machine would run 4 instances at the fastest speed - but maybe someone who regularly runs multiple back tests could give you a better answer.


RE: What hardware is adequate for tick backtesting? - llforex - 05-27-2015

(05-27-2015, 10:05 AM)d4v3 Wrote: You are not allowed to view links. Register or Login to view.As I said previously I would think that you will get the maximum speed by running one MT4 application per core - so a quad core machine would run 4 instances at the fastest speed - but maybe someone who regularly runs multiple back tests could give you a better answer.

Got you.  Many thanks, d4v3.  Only 4 simultaneous running of mt4 backtestings.  I will check into servers.

Merci beaucoup.

llforex