Run EA while markets are closed for testing code ?

New comment
 
Is there anyway to run an EA other than the stratagey tester while the market is closed to do test runs or code that I write?
 

strategy tester simulates an open market, so your question doesnt makes any sense.

the only way to test an ea while market is closed, is simulation of an open market...

 
meikel:

strategy tester simulates an open market, so your question doesnt makes any sense.

the only way to test an ea while market is closed, is simulation of an open market...

Not entirely true, in the sense I think you're intending it. For comparison, the only thing I really miss from NinjaTrader is the simulated price feed with a slider for influencing which way the simulated prices go. Dramatically reduces the difficulty of testing that things like trailing stops are working correctly. In practical terms, that kind of simulated price feed is different to backtesting using historic data.

  


 
Experimental samples, that can work in both live and test modes while the market is closed:
 
===============================================================================================================
2010.02.27                            Programs  Locations  Size, kB    LOC           LOD              Downloads
---------------------------------------------------------------------------------------------------------------
1. A System: Original                               ru en       245  15753
2. A System: Code Sample 1                             en        23    792
3. A System: Code Sample 6                             en       188   5588
---------------------------------------------------------------------------------------------------------------
Total:                                       3          4       456  21133
===============================================================================================================
 
Designations:
LOC - lines of code
LOD - lines of documentation


 
It's most definitely NOT a 'meaningless question', though it may be so for you. If you can't see the validity of it, or have no need of it, that doesn't mean it isn't to others that would find it to be a very useful and productive tool if a reliable, accurate simulator were available.

I certainly hope that they do a much better job of it in MQL5.

I agree with you that the 'strategy tester is a VERY poor substitute, and in fact is outright worse than nothing at all in many cases both for testing code comparing results and testing and running the code in the same manner as it is when the market is active. An accuracy rate of 25% ~ 50% is worse than having nothing at all because it only leads to invalid, meaningless conclusions that are little better than guessing.

GIGO

 
Ais:


What are theses test results from? What program, utility, script etc are you using to generate these?

Does it provide more functionality and accuracy that the strategy tester in MQL4 ?

 
meikel:

strategy tester simulates an open market, so your question doesnt makes any sense.

the only way to test an ea while market is closed, is simulation of an open market...

You are correct, which is exactly why and what I am looking for.

MQL4's strategy tester is so flawed and inaccurate that it can not really be classified as 'simulating an 'open market''

It does not operate, act and react anywhere near like an EA running on a chart while the market is open and active.

 

FourX wrote >>

MQL4's strategy tester is so flawed and inaccurate that it can not really be classified as 'simulating an 'open market''

It does not operate, act and react anywhere near like an EA running on a chart while the market is open and active.

The Tester has it's limitations and flaws, but a lot of them can be overcome by various techniques or the inaccuracy can be measured to a certain degree. It's far from perfect, but if used properly and carefully can be a powerful tool.

 
FourX:

An accuracy rate of 25% ~ 50% is worse than having nothing at all because it only leads to invalid, meaningless conclusions that are little better than guessing.

I assume u r referring to 'Modeling quality' - that has absolutely nothing to do with the accuracy of Testing results.

 
gordon wrote >>

I assume u r referring to 'Modeling quality' - that has absolutely nothing to do with the accuracy of Testing results.

I what way Gordon? How can the results be valid if it is using a modle that it is using and basing it'soperation is only 25% accurate? It is inherently going to be way off because it is using the data from the modeling qulity that is ~ 75% innacruate ?

 
FourX:

I what way Gordon? How can the results be valid if it is using a modle that it is using and basing it'soperation is only 25% accurate? It is inherently going to be way off because it is using the data from the modeling qulity that is ~ 75% innacruate ?

I suggest u study about 'Modeling Quality' and how it's calculated here -> Articles -> Tester -> One-Minute Data Modelling Quality Rating


Having 25% 'Modeling Quality' does not mean that "it's operation is only 25% accurate" and that "the data from the modeling qulity that is ~ 75% innacruate". You are miss-interpreting this figure's meaning. The figure only has meaning within the 'time-frame' paradigm... Which is mostly only relevant for TA based experts and cannot be used as a general figure. Furthermore, this figure is a 'weighted' figure, and so is inherently biased.

12
New comment