Hedge and Correlation Strategy | Page 205 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Correlation + Momentum Strategy? 6 replies

MT4 MTF Correlation table and Correlation oscillator indicators 155 replies

InsomiaFX Correlation Double Hedge EA 66 replies

Looking for EA for Correlation Strategy 6 replies

Hedge strategy (not a hedge fund) 13 replies

Attachments: Hedge and Correlation Strategy
Exit Attachments

Hedge and Correlation Strategy

  • #4,081
  • Jan 27, 2015 12:49pm Jan 27, 2015 12:49pm
Quoting ChuckNZ
The following relates to the ZScore 3-pair EA only. I carefully reviewed the M15 data available at MATAF and came up with the following sets that I have been running for the last 24 hours. Only one trade (AUDCAD/USDCAD/AUDUSD) and it was profitable. Very little deviation from the normal correlation in the rest of the sets. I'm trading all nine of these sets on 8 live accounts plus one demo account. The demo account is only so that I can see the action on my own PC. The live accounts all run on VPS's. Anyone else trading this EA? I think it may be...
Please give me the your AUDCAD/USDCAD/AUDUSD settings..TAnk you.
  • #4,082
  • Jan 27, 2015 6:50pm Jan 27, 2015 6:50pm
Quoting efraim77
{quote} Please give me the your AUDCAD/USDCAD/AUDUSD settings..TAnk you.
Just the default settings except for:

Pair2 USDCAD
Invert2 True
Pair3 AUDUSD
Invert3 False
  • #4,083
  • Jan 27, 2015 6:52pm Jan 27, 2015 6:52pm
ZScore 3-pair has a serious problem. If any of the orders fail (off-quotes or whatever), it doesn't try again. This leaves you a high risk of loss. I will fix and upload it later today. I will use the OrderReliable library, which will be uploaded as well.
  • #4,084
  • Jan 27, 2015 10:03pm Jan 27, 2015 10:03pm
Ok, here is the fixed EA and the OrderReliable library needed to ensure that all trades are executed. Just put OrderReliable_120205.mqh in your Include folder. I suggest that you use the OrderReliable library in all of your EA's. The documentation at the front of the code tells you all about it. It's been used by hundreds of traders over the years and is rock solid.

The reason for this change is I caught the EA just putting on two legs of the trade after the third leg failed. Very risky.

If you look at the change log at the front of the EA you will see all of the changes that I have made:

// Version 007
// Change TP to be dollars (or pounds or euros) rather than pips
// Change TP to be double
// Modify MagicNumber to reflect number of bars (Period_) and timeframe (Period()) can now run EA on two timeframes
// Display MagicNumber
// Don't start trading until ZScore is below RestartLevel (default = 2.0)
// After hitting basket TP, don't take new trades until ZScore is below RestartLevel
// Display number of bars (Period_) and timeframe (Period())
// Can still run five levels, but change default to only trade three levels
// Make sure we have sufficient bars for all three pairs (was only doing this for two pairs before)
// Version 008
// Show "(i)" if pair is inverted
// If Pair2 or Pair3 are not valid symbols, tell user
// Version 009
// Interface to OrderReliable library to prevent failed orders
Attached File(s)
File Type: ex4 ZScore Diff 3 pairs EA_009.ex4   82 KB | 609 downloads
File Type: mq4 ZScore Diff 3 pairs EA_009.mq4   14 KB | 886 downloads
File Type: mqh OrderReliable_120205.mqh   64 KB | 1,439 downloads
  • #4,085
  • Jan 27, 2015 10:38pm Jan 27, 2015 10:38pm
Oops... I just found another serious bug in the original code for ZScore 3-pair". Line 128 was moving a value to "C2", then using the value in "C3". I don't understand how it has been working as well as it has. One always wonders whether it's a bug or a feature. Hopefully, the fixed code will outperform the broken code. I'm running live on 8 accounts. No losses yet, in spite of bugs.

A bit of a tip.... nothing is going to happen until the ZScore (displayed on the chart) reaches 2.5. It can take minutes, hours or days.

Remember that these later versions use the OrderReliable library that was uploaded in one of my earlier posts.

Another unsolicited tip.... I'm going to download the spreadsheet from MATAF once a week and use only the pairs with correlation > 85 or < -85.
Attached File(s)
File Type: ex4 ZScore Diff 3 pairs EA_010.ex4   83 KB | 662 downloads
File Type: mq4 ZScore Diff 3 pairs EA_010.mq4   14 KB | 913 downloads
  • #4,086
  • Jan 28, 2015 2:43am Jan 28, 2015 2:43am
I am very grateful for the work you doing with EA. It's very deep analyse, I'm learning a lot.
  • #4,087
  • Jan 28, 2015 3:06am Jan 28, 2015 3:06am
Quoting Pierre
{qnot necessarily forming a neutral rangeote} Hi ChuckNZ the 3 pairs Hedge was a system I tried on the past for example in the following pairs correlation is not important because the trade is neutral but the spread of the 3 pairs will be your loss S EURUSD B GBPUSD B EURGBP or B EURUSD S GBPUSD S EURGBP I will continue with the original Zscore 2 coupled pairs
But you can use any 3 pairs,
  • #4,088
  • Jan 28, 2015 3:06am Jan 28, 2015 3:06am
Quoting ChuckNZ
{quote} Just the default settings except for: Pair2 USDCAD Invert2 True Pair3 AUDUSD Invert3 False
Thank You!
  • #4,089
  • Jan 28, 2015 12:48pm Jan 28, 2015 12:48pm
Some questions for those of you who understand better than I do about ZScore and how we are trying to use it.

I watched six sets running for 14 hours with the corrected code and the ZScore didn't get much higher than 0.5 on any of them. Do you still think 2.5 is a good trade level or did we end up there due to the bug that I fixed on line 128 of the code?

I provided a list of several sets of three, but I think that I screwed up.

Earlier discussions, for instance, talked about a set like this:

EURUSD
GBPUSD
EURGBP

The sets mentioned had what I call "symmetry". Each currency (EUR, GBP, USD) appeared twice.

Some of the sets that I suggested didn't have that symmetry. For instance, I suggested:

EURJPY
EURGBP
EURUSD

While all three are highly correlated to each other, the symmetry has vanished. Is it valid to use a set like that or was I wrong to suggest it?

By the way, if you aren't interested in the source (.mq4) and only use the .ex4, you don't need to be concerned about the OrderReliable library. It is already compiled into the .ex4.

Good trading!
  • #4,090
  • Jan 28, 2015 9:45pm Jan 28, 2015 9:45pm
Another update of ZScore 3-pair for you. Changes in this version include:

// Version 011
// Display high and low water mark for ZScore
// Display high and low water mark for Profit
// Display TP


I'm uploading the source and the ex4 plus the OrderReliable library. If you are only using the ex4, you don't need to worry about the library.

I continue to get small, steady profits with the following pairs although they are less correlated now than they were last week:

AUDCAD
USDCAD (inverted = true)
AUDUSD

Some feedback or dialogue would be useful.
Attached File(s)
File Type: ex4 ZScore Diff 3 pairs EA_011.ex4   85 KB | 841 downloads
File Type: mq4 ZScore Diff 3 pairs EA_011.mq4   16 KB | 1,519 downloads
File Type: mqh OrderReliable_120205.mqh   64 KB | 1,192 downloads
  • #4,091
  • Jan 28, 2015 10:27pm Jan 28, 2015 10:27pm
Quoting ChuckNZ
AUDCAD USDCAD (inverted = true) AUDUSD Some feedback or dialogue would be useful. {file} {file} {file}
Hi Chuck,

I can run some analysis in R (an open source statistical analysis program) to test various parameters and settings for this trading idea. I am new to the thread, but toying with some ideas of trading co-integrated baskets. Before I set this up to analyze in R, first a few questions:

1) What timeframe will this run on?
2) Are you buying/selling equal amounts of each pair?
3) For the example above: AUDCAD, USDCAD(inverted=true), AUDUSD.
3a) Which pairs do you buy/sell if the Z score is -2?
3b) Which pairs do you buy/sell if the Z score is 2?

I may think of more questions later.

EZcurrency
  • #4,092
  • Jan 28, 2015 10:47pm Jan 28, 2015 10:47pm
Quoting ezcurrency
{quote} Hi Chuck, I can run some analysis in R (an open source statistical analysis program) to test various parameters and settings for this trading idea. I am new to the thread, but toying with some ideas of trading co-integrated baskets. Before I set this up to analyze in R, first a few questions: 1) What timeframe will this run on? 2) Are you buying/selling equal amounts of each pair? 3) For the example above: AUDCAD, USDCAD(inverted=true), AUDUSD. 3a) Which pairs do you buy/sell if the Z score is -2? 3b) Which pairs do you buy/sell if the Z...
Great to hear from someone delving into this area. I may not be the best person to answer your questions, but I'll give it a try. A trader by the name of "Ronaldo" wrote the EA. I just spruced it up and fixed a few bugs. But, I'm working on it full time, so that probably qualifies me (LOL).

The EA will work in any timeframe. It was suggested to me that I should use M15 and that's where I have put my efforts.

Yes, the EA buys/sells equal lot sizes, probably less than ideal. There's been a bit of discussion here about better sizing methods.

I tested the EA to answer your 3rd question to make sure I give you the right answer. Keep in mind that in the example you quoted, USDCAD is inverted. Here is what happens:

ZScore > 2.5:

Sell AUDCAD
Sell USDCAD
Buy AUDUSD

ZScore < 2.5:

Buy AUDCAD
Buy USDCAD
Sell AUDUSD

I'm frustrated trying to find some sets of three pairs that are highly correlated at the moment. Using 200 bars of M15 data, there aren't any.

I'm going to try 100 bars to see if I can get any action. I'm using the data at MATAF to come up with sets to trade.
  • #4,093
  • Jan 28, 2015 11:03pm Jan 28, 2015 11:03pm
Quoting ChuckNZ
{quote}ZScore > 2.5: Sell AUDCAD Sell USDCAD Buy AUDUSD
Thanks for the quick reply. Let me make sure I understand correctly ... so if Zscore > 2.5, then:

Sell AC
Sell UC
Buy AU

So more or less, AUD cancels out (one sell, one buy), and we are then net short USD and long CAD (so essentially short USDCAD).

Correct, or am I wrong since UC is inverted?

Regards,
EZ
  • #4,094
  • Jan 28, 2015 11:12pm Jan 28, 2015 11:12pm
Quoting ezcurrency
{quote} Thanks for the quick reply. Let me make sure I understand correctly ... so if Zscore > 2.5, then: Sell AC Sell UC Buy AU So more or less, AUD cancels out (one sell, one buy), and we are then net short USD and long CAD (so essentially short USDCAD). Correct, or am I wrong since UC is inverted? Regards, EZ
You are correct. There's been hundreds of posts about "Why not just short USDCAD and save some spreads?".

I don't have a good answer to that question, do you?
  • #4,095
  • Jan 28, 2015 11:22pm Jan 28, 2015 11:22pm
And to check, I just plotted some older 1 hr data I had stored in R. The black line is AUDCAD+USDCAD-AUDUSD, and red line is USDCAD.

I think they are different, since I accounted for the approximate pip value:
audcad$AUDCAD <- audcad$AUDCAD*0.91
usdcad$USDCAD <- usdcad$USDCAD*0.90

AU was $1/pip
Attached Image
  • #4,096
  • Jan 28, 2015 11:23pm Jan 28, 2015 11:23pm
I don't have a good answer to your question though.

I'll show you what I've been working on and you can tell me what you think ... I'll post shortly.

EZ
  • #4,097
  • Jan 28, 2015 11:43pm Jan 28, 2015 11:43pm
Quoting ezcurrency
I don't have a good answer to your question though. I'll show you what I've been working on and you can tell me what you think ... I'll post shortly. EZ
I might make a new version of the EA that just takes the single position and compare the results. The proof, as they say, is in the pudding. I can't answer the question otherwise.
  • #4,098
  • Jan 29, 2015 12:16am Jan 29, 2015 12:16am
Chuck,

What do you think of this idea below? I have an R script to analyze this for two and three pairs. Below is outlined for two pairs ... 3 pairs is similar, but there is something wrong with my script so I can't produce any analysis now ... I'll look at it later.

For two pairs ... we are looking for co-integrated pairs. This is different than correlated. Cointegrated means that the difference between the spread of the two is mean reverting. So for example if you have 1 unit long of of EURGBP and short 1.3 units of NZDCAD, if you plot the difference between the two you get a graph like the one below.

The black line is the spread (so 1 EURGBP - 1.279 NZDCAD). This is long EG, short NC when buying the spread. Short EG, long NC when shorting the spread. The red lines are -1 and -2 stdev, and green are +1 and +2 stdev. The idea is to buy the spread when at -1 or -2 stdev, then exit at the mean (black horizontal line). Vice versa above for shorts. Another variation is buy the spread at -1 stdev, then buy twice as much at -2 stdev, and exit at mean horizontal line.

This is analyzed on some old 1 hr data, but we can use any timeframe. I think larger TF's are better. I ran a linear regression of each currency pair (about 25) on all other pairs, calculated the spread, then ran an Augmented Dickey Fuller (ADF) test to test if the spread was statistically mean reverting. The lower p-value means its more mean reverting. I only kept p-values < 0.05. This resulted in about 100 sets of two pairs which are co-integrated. I will post more graph samples later.

Perhaps you can code this into an EA? Not the analysis ... I can do that ... but make the EA buy (or short) A units of currency X, and short (or buy) B units of currency Y when the spread (which the EA can calculate) is below a level which is user inputted.

EZ
Attached Image
  • #4,099
  • Jan 29, 2015 12:20am Jan 29, 2015 12:20am
More sample cointegrated pairs
Attached Images
  • #4,100
  • Jan 29, 2015 2:44am Jan 29, 2015 2:44am
Quoting ezcurrency
Chuck, What do you think of this idea below? I have an R script to analyze this for two and three pairs. Below is outlined for two pairs ... 3 pairs is similar, but there is something wrong with my script so I can't produce any analysis now ... I'll look at it later. For two pairs ... we are looking for co-integrated pairs. This is different than correlated. Cointegrated means that the difference between the spread of the two is mean reverting. So for example if you have 1 unit long of of EURGBP and short 1.3 units of NZDCAD, if you plot the difference...
I have to tell you that I'm having a bit of trouble understanding what you want to do. I'll get there in the end, but struggling a bit right now.

You say "spread" and I'm thinking the difference between bid and ask and I can't relate that to what you are doing. By "spread", are you referring simply to the difference in (say) bid prices of the two pairs?
0 traders viewing now