Hello,

I have a few questions and will may it brief.

My background is in statistics and I have written all my trading programs from scratch (proto-type Python/R and execution C++) I can not by nature use black box programs, I need to re-create the processes independently. I noticed that the "Z" scripts are executable files and there has been reference to being able to recreate them by looking at the tutorial.

Two questions:

1. Has anyone done that for Z7? It's a machine learning algorithm. Where you successful?

2. It is trading in demo now and has been. No trades yet. Has anyone gotten this to work.

I am sure many people in this forum are like me and have spent enormous amounts of time chasing rabbits done the rabbit hole, only to find later that the only thing accomplished was to move the calendar forward 3 months. Trying to avoid repeating my mistakes.

Thanks in advance

Only few trades on my z7 live on real Oanda account. this is the log:
Trade Z7 portfolio

Bar period 1 hour (avg 86 min)
Trade period 01.08.2016-15.08.2016

Gross win/loss 28$ / -23$ (+18p)
Average profit 135$/year, 11$/month, 0.52$/day
Max drawdown -54$ 1063% (MAE -55$ 1075%)
Max down time 47 hours from Aug 2016
Max open margin 80$
Max open risk 93$
Trade volume 23994$ (639142$/year)
Transaction costs -3.68$ spr, 0.16$ slp, 0.00$ rol
Capital required 242$

Number of trades 4 (107/year, 3/week, 1/day)
Percent winning 75.0%
Max win/loss 18$ / -23$
Avg trade profit 1.27$ 4.5p (+32.5p / -79.6p)
Avg trade slippage 0.04$ 0.1p (+0.5p / -1.1p)
Avg trade bars 6 (+7 / -6)
Max trade bars 10 (10 hours)
Max open trades 2
Max loss streak 1

Annual return 56%
Profit factor 1.22 (PRR 0.26)

Portfolio analysis OptF ProF Win/Loss Result

EUR/USD avg .1000 ++++ 1/0 +5$
GBP/USD avg .1000 0.81 1/1 -4$
USD/JPY avg .1000 ++++ 1/0 +4$

PA avg .1000 ++++ 2/0 +9$
PG avg .1000 0.81 1/1 -4$

EUR/USD:PA .1000 ++++ 1/0 +5$
EUR/USD:PA:L .1000 ---- 0/0 +0$
EUR/USD:PA:S .1000 ++++ 1/0 +5$
GBP/USD:PG .1000 0.81 1/1 -4$
GBP/USD:PG:L .1000 ---- 0/0 +0$
GBP/USD:PG:S .1000 0.81 1/1 -4$
USD/JPY:PA .1000 ++++ 1/0 +4$
USD/JPY:PA:L .1000 ++++ 1/0 +4$
USD/JPY:PA:S .1000 ---- 0/0 +0$

Thanks for your reply.

Have you tried re-creating the Z7 using the documentation that Zorro provides. Their suggestion seems to be that they are not hiding any algorithm and have provided enough information to make it your own. I was hoping to tweak the algorithm and try some other optimizations, but having some trouble getting up and going?

Any code as a starting point would be appreciated.

Thanks

Workshop 7: Machine Learning. Price Action Trading
&
http://www.financial-hacker.com/build-better-strategies-part-5-developing-a-machine-learning-system/

are the only info I am able to provide you.

Ciao

Very nice article.
However, I do not understand:
1. where is the 2016 script repository?
2. can you successfully run the script that generate the data set?
I have Zorro free and it really runs long time.

What is your experience?
Thanks for sharing.
Cheers

1. come on, I am sure you can find it
2. I have not tried. Otherson the forum are more machine learning oriented. Sorry

I tried it and it worked fine when carefully following all instructions. Don't remember the total run time but I believe it was less than 15mins (i7 2.8GHz CPU and lots of memory).

I could run it as well and I found the code repository.
It was so clear where the code is that I could not find it.
However, if I try to change the layer vector from:

c(50,100,50)

to:

c(50,80,80,80,50)

I get an error. I am not sure whz, I need to investigate more.

I wonder to know whz this strategy does not work with GBPUSD?
and another thing, I can't find the function scale() in the manual.
So wath's the meaning of this?

var change(int n)
{
return scale((priceClose(0) - priceClose(n))/priceClose(0),100)/100;
}

anyone working on this?

http://www.zorro-trader.com/manual/en/norm.htm

But I think there is indeed no link to it from the Site Map, it is only reachable by "What's new" -> "Nomalization functions".