MT4 Strategy Tester Computer Hardware Bottleneck | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Trading computer set up (hardware) 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 Computer Hardware Bottleneck

  • #1
  • First Post: Sep 16, 2011 1:34am Sep 16, 2011 1:34am
Why does it take MT4 Strategy Tester so long to run a back-test of an EA? I understand that the amount of time required is dependent on the complexity and the quality of the EA programming, but this is ridiculous.

Where in the computer hardware would be the bottleneck? If cost wasn't an issue, build for me your dynamo computer system to blaze through strategy tester!!
  • #2
  • Sep 16, 2011 2:49am Sep 16, 2011 2:49am
If you are using version 406 then you are probably seeing poor performance due to a bug . . .

Other than that you need a fast CPU, and reasonable disk access performance.

CPU, something like the i5 2500k and overclock it to 4GHz

Maybe your EA can be optimized for better performance ? don't recalculate on every tick if doing it once per bar will suffice . .
20 pips a day isn't too much to ask . . .
  • #3
  • Sep 16, 2011 5:59am Sep 16, 2011 5:59am
Hey RaptorUK, thanks for the reply. I am using version 4.0 build 225.

I am using an I currently am running a core i5 - 760 processor (2.8ghz) with 4 cores and 8mb cache. I have two SSD's running in RAID 0 (this is giving me around 550 mb/s read and 411 mb/s write speeds). I don't know much about it but I am starting to wonder if the bottleneck has something regarding terms like: I/O transfer capability, bus spped, DIM (intel term).
  • #4
  • Sep 16, 2011 6:27am Sep 16, 2011 6:27am
Quoting bigsteely101
Hey RaptorUK, thanks for the reply. I am using version 4.0 build 225.

I am using an I currently am running a core i5 - 760 processor (2.8ghz) with 4 cores and 8mb cache. I have two SSD's running in RAID 0 (this is giving me around 550 mb/s read and 411 mb/s write speeds). I don't know much about it but I am starting to wonder if the bottleneck has something regarding terms like: I/O transfer capability, bus spped, DIM (intel term).
PC speed won't make much difference to the tester as it's only a single threaded process and can't take advantage of the better hardware. It is only using one of your cores and doesn't write enough data to warrant faster disks.

MT5 is far superior in the testing department allowing you to share the process between multiple cores or even other PCs.

Depending on the EA you're testing, you could use the control points model in MT4 (rather than the every tick mode) to get close to your parameters and then use every tick to fine tune it with fewer parameter options. Reduce the number of steps as much as possible.

Having said all that, the test results produced will be nowhere near the results you get running live and I would suggest you just run it on demo or, better still, a micro live account using cents instead of $$. The tester is only really perhaps fit for proving the logic of an EA and is a waste of your time if you expect the EA to produce the same results when it goes live.

Regards, Paul.
  • #5
  • Sep 16, 2011 6:55am Sep 16, 2011 6:55am
Quoting RaptorUK
If you are using version 406 then you are probably seeing poor performance due to a bug . .
I installed another instance of Mt4 but this time updated the build to version 402. I just saved myself 82 hours of testing!! The build 225 is estimating 210 hours, whereas the build 402 is saying 128 hrs. Not great but definitely an improvement. Thanks guys for the help. If anyone else has any suggestions feel free to post them. Bringing it down another 40% would be sweet.
  • #6
  • Sep 16, 2011 7:07am Sep 16, 2011 7:07am
Quoting pah
Depending on the EA you're testing, you could use the control points model in MT4.

Having said all that, the test results produced will be nowhere near the results you get running live and I would suggest you just run it on demo...
Regards, Paul.
I am testing between 2008/09/08 and 2009/01/08 with control points and I still have 128 hrs to wait it out.

I couldn't agree with more. It's absolutely imperative to trade/test your EA on live demo trading to see if it works. I made the noobie mistake and started looking at villas in Bali that I was ready to buy with the expected windfall I was going to encounter!! LOL
  • #7
  • Sep 16, 2011 7:20am Sep 16, 2011 7:20am
Quoting pah
PC speed won't make much difference to the tester as it's only a single threaded process and can't take advantage of the better hardware.
If the EA is using one of the cores at 100% CPU then using a faster CPU will certainly improve things . . .
20 pips a day isn't too much to ask . . .
  • #8
  • Sep 16, 2011 7:29am Sep 16, 2011 7:29am
Quoting bigsteely101
Hey RaptorUK, thanks for the reply. I am using version 4.0 build 225.

I am using an I currently am running a core i5 - 760 processor (2.8ghz) with 4 cores and 8mb cache.
It's most likely to be your CPU . . . look at Task Manager, then processes, sort by CPU so the Process with the highest CPU use is at the top. If you have Hyperthreading enabled your Terminal is going to use a max of around 18% from what I have seen . . . this will be maxing out one core.

It sounds like you are doing an Optimization, if you are optimizing by many parameters then the run time will get long . . .
20 pips a day isn't too much to ask . . .
  • #9
  • Sep 16, 2011 8:43am Sep 16, 2011 8:43am
My average CPU % is 24. This is for the one that I am running an optimization on. I am also using 7 threads. The other four instance s that I am running are between 3 and 5%. Thanks for the help
  • #10
  • Sep 16, 2011 8:51am Sep 16, 2011 8:51am
You weren't really clear about this in your PM, but are you running a backtest? or an optimization?

MT4's optimizer is a brute-force backtester. It just runs the backtest once for every permutation, so optimization time = backtest time * # Permutations.
  • #11
  • Sep 16, 2011 4:53pm Sep 16, 2011 4:53pm
Quoting Ronald Raygun
You weren't really clear about this in your PM, but are you running a backtest? or an optimization?

MT4's optimizer is a brute-force backtester. It just runs the backtest once for every permutation, so optimization time = backtest time * # Permutations.
I'm running both, back-test and optimizer, at separate times though.


That is ridiculous! You would think that MT4 developers could optimize their optimizer to run better.
  • #12
  • Sep 16, 2011 9:01pm Sep 16, 2011 9:01pm
Quoting bigsteely101
I'm running both, back-test and optimizer, at separate times though.

That is ridiculous! You would think that MT4 developers could optimize their optimizer to run better.
They did, but it's in MT5...

Making the sort of changes necessary to improve performance significantly under MT4 would have required such a big re-write that it wouldn't be worthwhile. It's natural that it's been done in the next version.

Regards, Paul.
  • #13
  • Sep 16, 2011 9:57pm Sep 16, 2011 9:57pm
Quoting pah
They did, but it's in MT5...

Making the sort of changes necessary to improve performance significantly under MT4 would have required such a big re-write that it wouldn't be worthwhile. It's natural that it's been done in the next version.

Regards, Paul.
Okay well maybe I should be doing my testing in MT5 then. Can I just load my existing indicators, EA's and scripts into MT5 like I would into MT4?
  • #14
  • Sep 17, 2011 2:26am Sep 17, 2011 2:26am
Quoting bigsteely101
Can I just load my existing indicators, EA's and scripts into MT5 like I would into MT4?
Nope, you will probably have to a lot of re-writing to get them to run in MT5.
20 pips a day isn't too much to ask . . .
  • #15
  • Sep 17, 2011 3:26am Sep 17, 2011 3:26am
Thanks RaptorUK for clarifying that.


Quoting pah
They did, but it's in MT5...

Making the sort of changes necessary to improve performance significantly under MT4 would have required such a big re-write that it wouldn't be worthwhile. It's natural that it's been done in the next version.

Regards, Paul.
What are you talking about? Are you working for THEM!! Don't most software companies make it a goal of theirs to provide a product/service that can retain previous customers. Let's see: Microsoft, apple, adobe, AutoCAD......the list goes on. If a company wanted to provide a legitimate service to traders they easily would have made it compatible. Why would any company have a business model that would cause them to lose all their existing customers by updating software. Doesn't make sense. Just one more of their dirty little tricks.
  • #16
  • Sep 17, 2011 9:33am Sep 17, 2011 9:33am
Quoting bigsteely101

What are you talking about? Are you working for THEM!! Don't most software companies make it a goal of theirs to provide a product/service that can retain previous customers. Let's see: Microsoft, apple, adobe, AutoCAD......the list goes on. If a company wanted to provide a legitimate service to traders they easily would have made it compatible. Why would any company have a business model that would cause them to lose all their existing customers by updating software. Doesn't make sense. Just one more of their dirty little tricks.
See, this is where you are misguided.

You are not the customer of Metatrader. Metatrader sells their platform to the broker, so they align themselves with the broker's interests.
  • #17
  • Sep 17, 2011 10:05am Sep 17, 2011 10:05am
Wow Ron you're as cynical as someone twice your age (like me).
You are in a maze of twisty little passages, all alike.
  • #18
  • Sep 17, 2011 11:28am Sep 17, 2011 11:28am
I suppose trading through two financial crises on different continents, dealing with a fund blowing up, and getting on that road to recovering capital that someone else lost gets to you. xD
  • #19
  • Sep 17, 2011 1:59pm Sep 17, 2011 1:59pm
Sorry, but how is it that I "get to you"?
  • #20
  • Sep 17, 2011 8:19pm Sep 17, 2011 8:19pm
Quoting bigsteely101
Thanks RaptorUK for clarifying that.

What are you talking about? Are you working for THEM!! Don't most software companies make it a goal of theirs to provide a product/service that can retain previous customers. Let's see: Microsoft, apple, adobe, AutoCAD......the list goes on. If a company wanted to provide a legitimate service to traders they easily would have made it compatible. Why would any company have a business model that would cause them to lose all their existing customers by updating software. Doesn't make sense. Just one more...
No, I don't work for them but as a professional programmer I can see why they would not want to take a perfectly stable (in most respects) version of a program and start heavily re-writing it. The time involved and risk of introducing new bugs far outweighs the alternative of adding the new features to a subsequent newer version.

Their re-write in this case is MT5 which is (code-wise) technically superior to MT4 and because of the new programming model, it will unfortunately require EA's and scripts, etc. to be programmed differently.

You want old code to run within a new environment which makes optimum use of newer hardware features, etc. Your expectations are somewhat unreasonable. Either you keep the old and slow code under the present system or you re-write it for the newer faster system.

Regards, Paul.
0 traders viewing now