The Great HGB Foward Test

Post by **trendinator** » Thu Jan 15, 2015 11:02 pm

thanks for your input boinvest, some great ideas there. will have to come back later tonight to reply properly, im very busy today.
didnt get your pm either, can you please re-send. thanks

qvintus · Post by **qvintus** » Thu Jan 15, 2015 11:19 pm

This is very useful on myfxbook for HGB. Under the Custom Analysis tab, click on trade and include comment as in the picture (for every trade type).
Then save each of them as a search profile. Name them Rad/Trend/Range/Wave/Trendwave. Then we can enable all trade types and filter to see exactly what we want.

https://www.myfxbook.com/portfolio/glob ... tb/1109327

Post by **trendinator** » Fri Jan 16, 2015 1:10 am

qvintus » Fri Jan 16, 2015 9:19 am wrote:This is very useful on myfxbook for HGB. Under the Custom Analysis tab, click on trade and include comment as in the picture (for every trade type).
Then save each of them as a search profile. Name them Rad/Trend/Range/Wave/Trendwave. Then we can enable all trade types and filter to see exactly what we want.

https://www.myfxbook.com/portfolio/glob ... tb/1109327

nice. custom analysis, filter by comment, is this perfect.

this will mean we can test 5x the amount of parameters, if every HGB has all trade types enabled. otherwise we will need soooo many accounts to have only one trade type per instance. personally i am in favour of this method. means we will be able to test more combinations with less instances.

on second thoughts, i think one trade type per instance will be better. easier to compare. what if one trade type blows the account? then all others are wasted. gonna roll with boinvest's suggestions

gringoh · Post by **gringoh** » Fri Jan 16, 2015 1:19 am

Hi Guys,

A few years ago I have developed an Excel tool to analyze trades thanks to MTCrap account history.

I have just updated this tool if it can help you with your analyze.

This new version is generating automatically pivot tables with slicers which allow to filter your report with multiple options. You can analyze multiple account.

You can find this little contribution in this thread:

http://www.stevehopwoodforex.com/phpBB3 ... =27&t=4057

I won’t be part of this forward test next week as I am already testing multiple instances of my dashboard and I haven’t got more resources available, but I’ll consider to be part of it soon.

Cheers,
G

Post by **trendinator** » Fri Jan 16, 2015 9:39 am

boinvest » Fri Jan 16, 2015 1:19 am wrote:
We really need dedicated trade type accounts.

Multiple accounts doesn't mean more work. It means having data easier at our disposal.
We could greatly benefit from a structured approach.

Actually you do want to have the 'verify trade privileges' thing on myfxbook.
As it unlocks very usefull tools.

I suggest we make use of a cloud storage. Anyone have suggestions??
Then make a folder for each participant..
Example...
Boinvest\TestRun01\Platform01
And then Place the SET files in these folders. (And maybe a Modded version of the EA)
Make a General section, at which people can Download, Instructions, Myfxbook installer, and the EA, Indi, Template, helper scripts.

The candidate EA to use right now is version 2i.

We have to group the testers in logical teams.
Based on the Latency to the Test Server.
So a first step would be, to gather data of the machines we have available to work with.
Maybe use a online office, or google docs for this to gather info.

And come up with a strategic plan what to test.
I've bloated some of my ideas what to test in a PM to trendinator.
Won't repeat the (possibly incomplete) large list here.
Got to make sure, we do usefull tests, Don't like to wast mine and other peoples energy and resources.

As tempting as it might be.. I can't be the Projectleader of this succesfull collabrative project.
My hands are filled with the HGB development.
So, I am very pleased to let someone else do this task. Trendinator?
But I will help where I can, provide input, strategy etc..

Think I can be more usefull coding the EA, share ideas and convert some concepts in modded versions for the testers to try.

ok so to summarise and reply to the most relevant points above.

One trade strategy per instance (eg. RAD only) - i think its the smart option
One myfxbook account per strategy - sounds good. i trust you know that this is best

Cloud storage for setfiles - sure. google drive is my pick. i have already started to create an excel spreadsheet with all the settings. will use google docs too when needed

test candidate 2i - ok but 2n is released now!? should we wait a little longer for a suitable candidate?

latency to test server should be checked - ok but will it really affect results? guessing slippage/requotes come into play here? amazon vps is around 14ms latency to global prime.

i am happy to be the project leader. no problem!

Post by **trendinator** » Fri Jan 16, 2015 9:52 am

Leslie » Thu Jan 15, 2015 11:27 pm wrote:
drunka » Wed Jan 14, 2015 3:43 pm wrote: Broker = Global Prime, 10.000 USD demo account.
If US and not Global Prime- shouldn't participate?

you can still participate by opening a demo account with global prime. HGB is not designed for US brokers, so no point testing that

Regner · Post by **Regner** » Fri Jan 16, 2015 1:41 pm

Hi all

I'am in Home Computer (2)

FXe3 · Post by **FXe3** » Fri Jan 16, 2015 1:45 pm

When you say not designed for US Brokers, I think your saying, not designed to be FIFO and Hedging compliant? Some brokers, such as ATC work fine with HGB, opening and closing trades as HGB instructs while maintaining compliance on the backend. If you like your broker and they don't support HGB look at using a tool such as One4all. However, for testing purposes I think we all should use GP to keep that one variable constant.

trendinator » Fri Jan 16, 2015 9:52 am wrote:
you can still participate by opening a demo account with global prime. HGB is not designed for US brokers, so no point testing that

FXe3 · Post by **FXe3** » Fri Jan 16, 2015 2:01 pm

Have we thought about how to handle corrupted testing, such as improperly executed, or test that get interrupted (computer shutdowns etc). QC should be implemented in some fashion if we want the data collected to be valid. Do we need two individuals running the same set to verify validity of data collected. Project planning is critical to properly implement/manage/and controlling all the variables. I just tossing out ideas for thought.

Post by **boinvest** » Fri Jan 16, 2015 4:48 pm

Yes, we did go through the update numbers yesterday very quickly. Most recent version is fine..
I'm not sure about some new stuff put in.. but hey.. that why we need a test

I think that the lack of available platform isnt really the challenge here.
The challenging part is, setting up the strategy, to determine which tests are usefull.
And how to do something usefull with the data we gathered.

Thanks for suggestion, Gringoh thats something to look into. (TLDR yet

)

Well latency was a thingy, maybe over optimizing here.. on the H4 test, it won't matter that much.

Qvintis, that what I've ment. I didnt knew you can save the items in the search section of fxbook.
Still cant find it.. But imagine you have to open 28 browser windows.
And fill in the filter like you did, in the screenshot.. Then go through the stats of all 28 accounts.
Look for rogeu trades, longest durations, best entrances, exits.. analyize biggest losers and wins. Use the advanced section et the end of history, to check for missed profit, drawdon ect.. Check trading hours entry eits.. and all other stuff we could analyze.. this is a massive job..
Plus I noticed, the filter doenst always work correctly..

But great, we decided to only have 1 type of trade per account

Yes Fx3e3, QC and planning is the key..

ok, well to take it to the next level..
Strange coming from my mouth as I thought I would never say this..
And it is my allergy...
it would be great to have some mathicians, statistic guys goe nuts over the data we are going to collect with Mathlab (or freeware equivalent)
Have somewhere read something about it with regarding to FX.