Robot Lab Project

Second version attached as requested.

New Inputs :

1. Time_Frame = 0; // Time Frame Used (0:current time frame)
   // 0:current time frame. Otherwize input number of minutes for desired time frame
   // 5 for M5, 60 for H1, 1440 for D1 and so on.
2. Max_Seconds_to_Activate = 60; // Max. Seconds to activate after candle starts

As I mentioned earlier, this robot need some filter, but I will leave that to you.

Should you find it profitable, kindly share your settings.

General Warnings

Few warnings regrading the Robots which are still in the lab and not approved (all at the moment)

1. Use only for back or forward testing
2. Robots are not optimized for Live Trading on Real accounts yet
3. Robots are not compatible with ECN brokers yet
4. 4-digit brokers use pips, while 5-digit brokers use points. a Point is 1/10 of a pip.
5. For input parameters in Pips (Pips is meant for 4-digit broker where 8=8 pips, multiple by 10 for 5-digit brokers, so 8 pips=80 points)

We are still at the beginning ...
Exploring a number of robots.
Optimizing them on sample Data.
Test them on larger data .
Then we'll filter them out and choose the best candidates for Demo trading.
...

Happy Birthday
Happy Newyear

I am currently getting prepared for some correlation robots.

To be able to do back testing on correlation robots (or any multi-pair robots), we have to switch to MT5.
So get prepared and install MT5 from your preferred broker (if you don't already have it ).

The first robot will trade divergence between correlated pairs.
Initially and to make things simple, the user will input TWO highly positively or negatively correlated pairs.

There are different methods on how to detect divergence between correlated pairs.

I like this one :

If the two pairs are positively correlated, the Robot will open virtual sell on one pair and virtual buy on the other, taking of course the pip value into consideration when deciding on Lot size for each pair, so that profit/loss comparison between the two orders makes sense.

Now As Profit/Loss percentage between the 2 virtual orders increases this means that divergence between the two pairs is increasing. As this percentage reaches some maximum figure (to be decided after testing), we open Actual orders.

Actual orders can be closed at certain pip profit or equity increase or when percentage of profit/loss for the virtual be at minimum.

Same thing goes for negatively correlated pairs. But this time the two orders will be of same type (either buy or sell)

This idea was inspired from insomiafx here. He's offline since June, 2013, I hope he doesn't mind.

Would like to know your opinions on this.

Correlation Divergence Trading (Two Pairs)

I though about this, and reached the following:

1. Lets take EUR/USD & GBP/USD as an example
2. Calculate Change Percentage [ 100 x (Close-Open)/Open] on the desired time frame for both pairs, so we have EU_Change% and GU_Change%
3. calculate Correlation Divergence CD = (EU_Change% - GU_Change%) x 100
4. Resulting CD values will oscillate below and above zero line, and can reach +100 or more and -100 or less
5. Actions based on CD Values:
   1. +50 to +80 : (Moderate Signal) Buy GU and Sell EU : Close when CD becomes Negative
   2. > +80 : (Strong Signal) Buy GU and Sell EU : Close when CD becomes Negative
   3. -50 to -80 : (Moderate Signal) Sell GU and Buy EU : Close when CD becomes Positive
   4. < -80 : (Strong Signal) Sell GU and Buy EU : Close when CD becomes Positive

   
   

I've attached the an image of the Excel sheet I'm working on. It worked perfectly for December.

The above is just brain storming in order to reach an applicable solution.
Above calculations over the Daily Time Frame made sense because Average Daily Range for both EUR and GBP is comparable with the resulting CD.

However, CD values calculated over different time frames need to be normalized over the selected time frame.
Also going down at smaller time frames we need to smooth CD values (through averaging ...)

Anyway, I'll have to spend some time on this before moving forward.
------------------------------

I just looked at the current price of EUR/USD and GBP/USD.
WOW, should we use the recommendation from the excel sheet in post # 206
by Selling GBP/USD and Buying EUR/USD we could be in good profit now at over +110 net pips on both positions,

So this correlated pairs hedging final return was over +160 pips in one day. (Attached Excel Image)



I know today was special. However I will develop the Robot based on this concept and see how it performs in back testing.

I quickly developed one using CCI, but nothing solid yet to post.

As I mentioned earlier we have to use MT5 for back testing multi-pair robots.
I don't know of you are aware of MT5 differences compared to MT4:

1. no-hedging on same pair,
2. Multiple trades are treated as a single position ...
3. SL and TP handling when you have multiple trades is not straight forward. ..

​just to name few.

But for the (Correlation_Hedger) robot its OK. As it will hedge different pairs.So stay tuned ...

PS: if the robot back testing results on mt5 were good, I can then develop an mt4 version for forward testing on mt4.

Correlation Divergence Trading (Two Pairs)

Corr_Diver_01 (MT5 Only)

So, here is the 1st version of this Robot, attached. Please note its for MT5 only. It will not run on MT4.

Input Parameters:

1. Magic_Number = 500001 // Magic Number
2. First_Pair = EURUSD // First Pair
3. Second_Pair = GBPUSD // Second Pair
4. Fixed_Lot_Size = 0.3 // Fixed Lot Size
5. Profit_EQ_Perc_Inc = 0.25 // Close at Profit as Equity%Increase
6. Loss_EQ_Perc_Dec = 1 // Close at Loss as Equity%Decrease
7. CCIPeriod = 2 // CCI period
8. tp_pips = 50000 // Take Profit Pips
9. sl_pips = 50000 // Stop Loss Pips

​This version used Commodity Channel Index (CCI) indicator, and traded divergence between CCI values on both pairs.


Test results for EUR/USD & GBP/USD on 2014

Full tester report attached as a compressed file.

Regards,

Corr_Diver_01 Robot

Test results with the same above inputs, from 2010 to 2015




Looks like we have something here ...

Corr_Diver_01 Robot

Test results for EUR/JPY and GBP/JPY for 2014. .... Interesting. For now use only for highly POSITIVELY correlated pairs.

Corr_Diver_01 Robot

Test results for EUR/JPY and USD/JPY for 2014. .... Interesting.

Very nice!

Going to test the idea and will upload results

Thanks fr the very valuable thread and all your hard work

Rooi

Tried to test myself but couldn't successfully download history data into platform. So unable to perform a full big period test.

Friend could you please do all the above tests from 2009? Thanks. I appreciate your work.
And also, what was a TF?
Thanks

Daily TF.

I tried on 2009 but History Data Quality was 0. So results are not reliable.

Hi Abokwaik

Please look at below journal screen shot and tell me if you know what I'm doing wrong.

I have never used the Mt5 Strategy tester before so dont know if it me but I doubt it

Thanks for your time

Rooi

Yor are using a 4-digit broker. So you have to decrease TP and SL by 1 digit.

Dear sir :
sorry for my ask...
the Corr_Diver_01 Robot have MT4 ver ??

Simple enough.

Thanks for the info.

Will upload my results

Rooi