Sending Offline Chart to Strategy Tester (25% model)
OfflineTestHelper is a simple script that helps with a simple testing strategies, nested in offline charts. The testing quality is displayed as 25%, though it highly depends on the original chart fidelity.
The Strategy Tester included in the Terminal only allows picking testing timeframes from a drop-down menu, which contains standard values M1, M5 etc. On the contrary, the offline charts never can have such timeframe value in a live terminal.
The script replaces one of the standard timeframes by the offline chart. It works with an artificial broker to prevent the refresh of such chart with on-line data.
Input Dialogue
Input lets you decide for alternative server name. It is useful, when you prepare multiple offline charts for testing, preserving all of them for later use. Otherwise keep the default.
The second option would systematically overwrite the candle open value by a close value of the preceding candle. Use this settings for offline charts with displacement of the open price, like a reversal candle on Renko charts has.
Result Dialogue
The First Steps
Watch the video from Theo on how to use the Helper with the offline chart.
What does the helper after dropping on a chart:
- Creates a virtual server folder. Absence of the on-line server will prevent the terminal from updating any chart.
- Copies essential Metatrader data to the new server folder (instrument list with latest quotes).
- Creates a copy of the current offline chart.
- Set its timeframe to M1.
- Removes zero-volume candles (they are not compatible with the tester).
- Stores the new chart into the virtual server folder.
Then it displays a result dialogue. You have a chance to stop the script now.
When you decide to stop, your data is prepared for use with any MT4 terminal instance. You may copy/move your virtual server folder to another terminal for testing, login into the virtual server in the other terminal.
When you decide to continue with the Restart, your current terminal instance will be used for testing. It will force the entire MT4 terminal restart.
- The terminal shuts down, unbinding current server data.
- Terminal starts with the virtual server “test”, the connection status remains “Disconnected”.
- Ignore whatever you see in the charts in front of you. Switch to Strategy Tester (ctrl+R) by hand.
- Pick symbol of your converted chart.
- Pick M1Â timeframe (your offline chart is stored as M1).
- Pick your strategy.
- Test.
- If you finished testing, log-in back to your original account&server (double click your previous account in the Navigator).
Sorry it was me again. Thanks for your help admin.
Do we have a release date for FXT Builder yet?
Well, should have been like a moth ago, I just did not make it. Most likely this Monday.
Would that be free to someone who has purchased OmniaBar chart package by any chance? 🙂
how to test renko more than 2048 candle ?i use off line test helper that cannot load more than 2048….
Your initial chart has to be as long as you require. In addition, the tester skips a bunch of initial candles in the chart, so you should increase the candles count by 1000 before sending the chart to the tester.
why are there difference of renko bar in tester and real account
i notice that in tester ,there are more renko bar than real account very much
example xauusd 15/07/2015 there are 2000 bar in tester, 30 bar in real account?
Hello kornlert, to be honest, I am not quite sure what you meant. If you are complaining about this script, it transfers exactly the same chart to the tester, it cannot add bars.
ON YOUR OFFLINE TEST HELPER, I RUN BACKTEST OF ANY EA AND I FIND THAT MY MODELLING QUALITY IS ABOUT ONLY 25%,,,,WHY?
AND HOW TO CORRECT
As mentioned above, the 25% is a standard accuracy for M1 chart testing. You may improve the quality by using FXT Builder instead. With the optional 3-rd party launcher you would get the 99% number displayed in the test report.
I found this tester very usefull in the past but it doesen´t seem to work anymore. Just tried on different mt4 platforms from diferent brokers and the result was about the same. Any ideas??
i just wish that it would not skip any bars but thanks for the scrip good work
Hello there ;
Thank you for this amazing script however when i tried to follow the guidelines provided here i got error message says ” Unsupported legacy hst version . Use b509 compatible script ” can any one help
The question was several times discussed, this script does not support chart formats used in Metatrader in 2013 or earlier.
this isn’t working. all the files are in the correct folder, so i proceed to follow the steps. however, when i get to the point where i want to click yes after bringing up the offlinetesthelper, the terminal restarts without bringing up the login box. it doesn’t allow to run it as test. it just goes to a demo offline chart when it pops back up without any bars
The login box is filled automatically, you do not need to fill it yourself in the recent versions.
hey everytime i put the script in and when the box pops up and says it will restart after i put test in the server. as soon as i click yes it restarts and doesn’t give me the chaance to put test in the server
The login box is filled automatically, you do not need to fill it yourself in the recent versions. It has been working like that for about two years.
Hello Milan, I am about to order the key for the renko builder and the backtest helper. Before I do that I wanted to share with you how I have tested the renko EA I have built so far.
I delete all charts from the terminal and also delete all history files. I only leave 1 min chart for the currency pair i want to test with the renko chart builder set to build the renko charts for 5 min time frame. After restarting the platform I ONLY open the offline chart 5 min and it is a pure renko chart. MT$ only downloads the history data for the time frame on the open chart. If I do not open the 5 min chart it only takes the data from the offline chart. I can also use the built in tester of MT4 to test the EA on 5 min charts.
Here is my question for you – we all know that MT4 does not store tick data and that is why we have to use 1 min chart. If I understand it correctly your renko chart builder buidls the renko charts on ticks, not on closing prices – is that correct? If that is so, the charts built from the live feed are built on ticks, but the charts built from history data are built on closing prices of 1 min- am I correct?
My second question – would there be a diference in the testing results between your backtest helper and my way of testing. If Yes – I would appreciate it if you tell me why?
Sorry for the lenghty post – I hope it did not annoy you.
Thanks and regards
Dear Milan, i have heard that downloading data from within the MT4 history center is not the best data to use for strategy testing or back testing, Would you agree with this ?
Will i get more accurate back test results if i download Tickstory and get the tick data from Dukascopy ? (which automatically converts it to FXT files)
When downloading the tick data do i only need to download the 1 min data ? or should i also download the 5min 15min 30min etc..? Or is there no need to get the other time
frames only the 1 min data ? As the back testing on the strategy tester is all done on 1 min.
Or if the modeling quality is only 25% regardless is it ok to use the history center data from MT4 ?
Thanks for your help.
Greg
Hi guys, I have the exact same question, as I can only get 25% modelling quality via MT4 data… If I use tickdata will this quality be improved once converted into renko bars?
This script has no ambitions, it takes ANY chart as it is, and copies its content to the M1 chart of a virtual ‘test’ broker. It also removes the zero volume candles (they would stop the backtester). Then the test is using the standard MT4 test engine. According to MQ definition of accuracy, the M1 testing reads always 25% regardless of its content.
If you are interested in tick based testing (aka 99% testing) of offline charts, you need to follow the link in the upper right corner to FXT Builders.
I had this working yesterday, and now today I get an error…
“Cannot call ‘Kernel32.dll::CreateFileW’, ‘Kernel32.dll’ is not loaded”
I have checked and I do have the ‘Kernel32.dll’ located in the libraries folder.
There is no other clue as to why this is happening. Any ideas? I am running windows 10 64bit, however as mentioned this worked yesterday without issues…
Note: I just tried the demo for RenkoFxtBuilder, and get the exact same problem… I also tried running the MT4 in administrator mode but it did not help…
OfflineTester doesn’t work with hybrid renko bars(( – (45000 zero bars….).
can it backtest W1 and MN1? thanks
Well, technically yes, though with 25% precision only. I would not recommend it.
I installed the offline tester in Scripts and followed the video but after I get the popup about zero volume bars , okay, and then MT4 restarts. But, it does not give me the log on screen. I pulled down; log into account and added the “test” . . . . still no bars. what am I doing wrong?
my MT4 is Oanda build 1090.
Suggestions ?
I’ve installed the OfflineTestHelper is not operating correctly, After I install the Helper onto the renko chart, the Input screen appears, I click OK.
The lengthy dialog box appears I click OK,…
Then, NO LOGIN box appears, MT4 simple shuts down and restarts with NO data in the chart and NO ‘WAITING FOR UPDATE’ message.
Build: 1090
Please help!!!!
You need the Strategy tester rather than charts.
hi Milan, I tried to download the 840 compatible offlineTestHelper_840.ex4, but it shows file not found. please check.thanks
Hello Simon, the 840 versions of indicators have been scratched after Metaquotes excluded the build 840 from the supported versions and from receiving the data feed. The live data feed is essential for those indicators to work properly.
Will this test give me similar results for a real account? I bought renko from you, but I have an ea test wrong ://
I am afraid this one is not suitable for EA performance proof, both the demo and real.
I wrote ea. I have 80% efficiency for the tester, but when ea works on the account, real efficiency is only 40%. How can I test ea on the tester correctly?