Hedge and Correlation Strategy | Page 127 | 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

  • #2,521
  • Feb 9, 2011 11:49am Feb 9, 2011 11:49am
gbpaud and gbpchf trades out with 22 pips profit.
Intraday only.
  • #2,522
  • Feb 9, 2011 11:50am Feb 9, 2011 11:50am
Quoting kishu0605
navk what is martingale strategy ? i guess i missed something
Google it. You will have detailed answer.
Intraday only.
  • #2,523
  • Feb 9, 2011 12:35pm Feb 9, 2011 12:35pm
Quoting navk
eurnzd i do not trade. Spread is more. Any way, what i noticed today triplets signals rarely comes. Quad signals come often. Why not enter at 95 or 98 for quad ? And if quad trades go big DD, then martingale strategy can be utised by adding more trades at some DD and have common profit target ?
Martingale is dangerous. However what Roundrock is suggesting isn't Martingale because it is limited by the Stoch Diff. Yes it can indeed end up in tears because it can treble the exposure and the risk but it has built-in limits.

I'm not advocating this but is worth testing.
  • #2,524
  • Feb 9, 2011 1:09pm Feb 9, 2011 1:09pm
i spoke about my backtests i done with a .dll.

here i give the mt4 version

i think you will understand how it works. simply put in the first and second pair.
then what stochastic difference criteria for entry and exit.
UseTimelessGraph set to true and you will see the equity like in the stratgy tester window. if false then you could compare the signals with exact entry and exit when it happens. put stoch diff indi for comparison.
invertfirst and invertsecond is the same like in the stoch diff indi.

load as many history data you can. i tested only with 1hour chart. try 30min or 4hour etc. the higher the better

now play with combinations and the invert switches
you will see if a paircombination is not went in profit then invert one pair or both. you will have four combinations for a hedge pair
try for example eurusd vs. gbpusd in 1h chart and with 80%entry 20% exit and switch invertsecond to true.

forgot to say that it take the pipvalue in calculations?

have fun with testing

Mediator
Attached File(s)
File Type: mq4 Equity Graph Go HomeV2.mq4   9 KB | 809 downloads
  • #2,525
  • Feb 9, 2011 2:45pm Feb 9, 2011 2:45pm
Quoting fortheex
I'm using 15TF with the rest 80% difference stoc entry point. Any commments?

Exit stertagy or rule ?
  • #2,526
  • Feb 9, 2011 3:03pm Feb 9, 2011 3:03pm
Quoting kishu0605
Hi,
can you please send me the indicator list of above chart ?

thank you & regards
kishu,

I used (at this time)
Stoch Diff 2 Pair (v2)
OffChartPrice
Stochastic different pairs 1.5b

Now they are not the recent ones, but -because of little time these days- I'm still using this configuration.

josch
trading is war - life is simple
  • #2,527
  • Edited 6:54pm Feb 9, 2011 6:36pm | Edited 6:54pm
Quoting Mediator
i spoke about my backtests i done with a .dll.

here i give the mt4 version

i think you will understand how it works. simply put in the first and second pair.
then what stochastic difference criteria for entry and exit.
UseTimelessGraph set to true and you will see the equity like in the stratgy tester window. if false then you could compare the signals with exact entry and exit when it happens. put stoch diff indi for comparison.
invertfirst and invertsecond is the same like in the stoch diff indi.

load as many history data you can. i tested...
Hi Mediator,

many thanks for this very nice equity backtester, well done!
I spent a few hours playing with it, made some changes (couldn't resist...) and i think i found something really interesting "by accident" -- while changing the code, testing various inputs combinations started giving suspicious results (too good to be true)...
I now need to find out where the bug is, or if it is my history data that is playing with me... or if it is simply the reality!
More later, I must get some rest now...

Here is what i mean by suspicious:
the figures on the right are the equity in USD for 1 Lot EURUSD / USDCHF, (that part of my mods i am sure about!), tested on M5 starting from 2010.12.01... +$40.000 in just 2 months... 877 trades (!) (i can explain why there are so many trades, but later)
All right, this doesn't count cost of spread on the 877 trades.

Attached Image (click to enlarge)
Click to Enlarge Name: equ.gif Size: 4 KB


Sq
  • #2,528
  • Feb 9, 2011 11:48pm Feb 9, 2011 11:48pm
Quoting squalou
Hi Mediator,

many thanks for this very nice equity backtester, well done!
I spent a few hours playing with it, made some changes (couldn't resist...) and i think i found something really interesting "by accident" -- while changing the code, testing various inputs combinations started giving suspicious results (too good to be true)...
I now need to find out where the bug is, or if it is my history data that is playing with me... or if it is simply the reality!
More later, I must get some rest now...

Here is what i mean by...
yes, this indi helps alot with testing. the good thing is, that you could backtest most every strategy with slightly changes in code.
but i'll be curios how you reached this results.

Mediator
  • #2,529
  • Feb 9, 2011 11:51pm Feb 9, 2011 11:51pm
Hi,
can anyone please tell me use of invertfirst symbol and invertsecond symbol ?
  • #2,530
  • Feb 10, 2011 12:43am Feb 10, 2011 12:43am
Hi guys,
Can anyone direct me to the post with explanation how to use stochastic differance to enter the trade for the pairs. I try for 2 days to find it but no succes.
Thanks.
Sasa

Quoting smjones
The person who changed my code did not get the array sizes correct. also they did not sort there currencies. Here it is sorted and corrected. Please use this indicator and this template.

I have no idea if the alerts work properly, because they have moved them to function calls and I did not follow the logic. So you will just have to test it.
  • #2,531
  • Feb 10, 2011 2:28am Feb 10, 2011 2:28am
Quoting Mediator
yes, this indi helps alot with testing. the good thing is, that you could backtest most every strategy with slightly changes in code.
but i'll be curios how you reached this results.

Mediator
Ok, i'm up after a few hours of sleep...
Thread was quiet during the night...


Yes, very interesting method indeed.
I like the "TimelessGraph" mode.
Clever way of plotting an equity curve!

I hope to be able to sched some light on my suspicious results today...
I've been thinking about it all overnight...
I think it has to do with the iStochastic() integrated function, which for some reason would not give reliable results:
I added some debug code to log every simulated "trade" found by the backtest indicator, along with the relevant stoch values that it used to enter/exit the trades, and when i look at them, and compare to a chart with both stochs loaded by the "Stochastic 2 Pairs sq" indicator, then i can see shifts in the values of the second pair (USDCHF):
the 2 stoch values logged by the Equity backtester and used for entering/closing trades are not from the same bar when i compare them to the stoch lines on the graphic indi: the UDSCHF stoch value is the one of 1 or 2 bars after the stoch value of the EURUSD...

NOTE: i have changed the K value of the StochDiff indicator to 72, to match that of the backtester indi (it's the first thing i noticed when loading the Equity backtester) so this is not the reason...

Now, which stoch value to trust ?
This is very strange, as both the Equity backtester indicator and StochDiff are based on the same iStochastic() function called with the same arguments, applied to the same history data...
There is absolutely no reason why they would give different values ?!?
The backtester uses the stochastic of the previous bar ("i+1") instead of the current bar, but that is absolutely expected: it simulates a trade taken at the opening of a new bar, so uses the values of the bar that just closed.
But it does it for both Pairs, so we should only see a shift of 1 bar for both pairs stoch values compared to the graphic, but not a shift between pairs.

This is a mistery so far, so i'll have to sched light on it before i can ever trust my results...


More to come...

SQ
  • #2,532
  • Feb 10, 2011 3:03am Feb 10, 2011 3:03am
good morning squalou,

i think it has to do with this line:

if (UseTimelessGraph)
{
backarray[trades] = Acc;
} else
{
---> PFBuffer[i+1] = Acc;
}

it only shift the graph 1 bar behind the signal. this i have done, because then i got a big spike at bar 0.

i comb through the code up and down, that i did not have a bug in there. but it is bug free (hope so). i use many more equity graphs for another systems and strategies. because i rate every equity of system if it is tradeable with the actual market situation or not. it is like filtering different systems.

Mediator
  • #2,533
  • Feb 10, 2011 3:48am Feb 10, 2011 3:48am
Quoting squalou
Ok, i'm up after a few hours of sleep...
Thread was quiet during the night...


Yes, very interesting method indeed.
I like the "TimelessGraph" mode.
Clever way of plotting an equity curve!

I hope to be able to sched some light on my suspicious results today...
I've been thinking about it all overnight...
I think it has to do with the iStochastic() integrated function, which for some reason would not give reliable results:
I added some debug code to log every simulated "trade" found by the backtest indicator, along with the relevant stoch...
Yes, I had this problem even with other Multi-currency IStochatic() functions. What I did to solve this problem is this.

PHP Code
bool run_once = true;

Init()
{
the init stuff here....
}

Start()
{
     if(run_once)
        {
             my code that would normally be in the start function goes here...
             run_once = false;
         }
     else 
         {
              the same exact code as above, but I do not use the index i 
              and  the for loop....
         }
This way the run_once paints the history and then by using a 0 in the index shift field of iStochastic(), The indicator cannot go back and change the totals when it gets bogus different info about the off chart pairs. I have found this problem in EVERY indicator that makes calls to off chart pairs. It seems to solve the problem
  • #2,534
  • Feb 10, 2011 3:49am Feb 10, 2011 3:49am
Quoting Mediator
good morning squalou,

i think it has to do with this line:

if (UseTimelessGraph)...
i don't think this makes any difference: the PFBuffer is what gets displayed on the chart, but it always takes its values from Acc, which is updated only when a trade closes.
Using PFBuffer[i+1] or PFBuffer[i] would only shift the curve by 1 bar, but not change the resulting values.

Ok, so for coders who would like to help in finding where the trap is, here is my "accident version" of the code i'm using right now to find it.

Steve, SMJ, Mediator, Roundrock, T1minator, others...
!!! YOUR HELP IS NEEDED HERE PLEASE !!!

I have hardcoded the default settings to the values i used to get the "fantastic" results i posted earlier, so you can all see by youself if it gives similar results or not.

Here are the settings (i changed the input names compared to Mediator's code, just to make them consistent with the other indicators we are using here):

extern string Pair1 = "EURUSD";
extern string Pair2 = "USDCHF";
extern int stochentry = 21;
extern int stochexit = 15;
extern double Lots1 = 1;
extern double Lots2 = -1;
extern datetime FromDate = D'2010.12.01';
extern bool UseTimelessGraph = true;
extern int K = 72;

Drop this indicator on an EURUSD M5 chart and see what it gives.

As you can see, the "accident" i am talking about is the "stochentry" value: 21 !!! we are very far from 80, 90, 95, etc...
And the smaller values will give the better results!!!

What it means is that we are somehow scalping while the stochastic diff is very low, so the stoch lines are close to each other, not widen like in the strategy.
And that seem to make a HUGE difference, unless my code is completely wrong.

Remember that the backtester simulates trades opened/closed at BAR OPEN ONLY, not every-tick; So it does not matter if the stoch reaches the open/close limits during the bar, it's only opening/closing at bar open with the stoch values of the previously closed bar.


We have to find out what my code actually simulates in terms of which pair is BUY and which is SELL (!), and then we can modify the EA to fit these conditions, and see if the EA gives the same results in forward testing than what the equity backtester reports...

You'll see i have added a 5 seconds delay after bar open in start(), i came up with this earlier this morning while studying the behaviour on an M1 chart with FromDate set to today (=D'2011.02.10').
Fior some reason MT4 seems not being updating its internal buffers consistently at a new bar open before a few seconds, giving inconsistent iStochastic() results (sometimes you get the same value as the previous bar, sometimes you get the new value....!?!)

Another note: i added logging capabilities to better analyse how iStochastic() behaves: if you set "log" to true, then it will create a CSV file containing all simulated trades and relevant information at each new bar.
It creates a different file for each new bar, with the name containing the date and time so you can compare the files contents and see if they give different results for consecutive bars.
And it DOES: this simply means that iStochastic() is giving me slighlty different values each time, although it shoudln't: past bars should not be changing, right ??? so should the stochastic, then.
You'll easily see what i meant with the bar shift explanations on my previous post if you look at several different files.
I used it on M1 chart to analyze this because it spits a new log CSV file every minute, so it's very quick to check.

Once we find the reason for this inconsistent stochastic behaviour, then we will be able to trust the results...

I only these could be real...

Sq
Attached File(s)
File Type: mq4 Equity Graph Go HomeV2 sq.mq4   9 KB | 683 downloads
  • #2,535
  • Feb 10, 2011 4:02am Feb 10, 2011 4:02am
Quoting smjones
Yes, I had this problem even with other Multi-currency IStochatic() functions. What I did to solve this problem is this.

[php]
bool run_once = true;

Init()
{
the init stuff here....
}

Start()
{...
thanks for that SMJ, that 's very interesting information...

I think i did something similar in my release, by waiting for each bar open, and i have even added a 5 more seconds delay.

I will try replacing the iStochstich() with a function that calculates it based on iHighest/iLowest() instead, and see if this makes any difference.

If not, then it means it is really the way mt4 reads the history data that is completely f***ed up !!!


Sq
  • #2,536
  • Feb 10, 2011 4:08am Feb 10, 2011 4:08am
Quoting squalou
thanks for that SMJ, that 's very interesting information...

I think i did something similar in my release, by waiting for each bar open, and i have even added a 5 more seconds delay.

I will try replacing the iStochstich() with a function that calculates it based on iHighest/iLowest() instead, and see if this makes any difference.

If not, then it means it is really the way mt4 reads the history data that is completely f***ed up !!!


Sq
That will not work. the problem is not iStochastic() It is that the History for the off chart currency changes in mt4 when it picks up mis-counted M1 bars. It is correct at the moment it happens, but can become incorrect in the history or even change several times. The only work around this that I found was accurate is to draw the history once and then only draw the current bar there after .Note that the current bar becomes the next bar in history when it is completed and cannot be revisited with this scheme.

And yes MT4 History on off chart currencies is completely fucked up!
This is one of many reasons, why I develop on MT4, but would never trade with size on Metatrader.
  • #2,537
  • Feb 10, 2011 4:09am Feb 10, 2011 4:09am
Some thoughts not backed up by EAs ...

Ref to the different exits it's unquestionable, that the stoch-problems are the same in OB- and OS-levels. If hedge pair is widening and prices remain at this level, stoch will go down and trigger exits.

If you suppose, that correlations really trying to 'go home', it would be helpful for better results to let such a trade running despite of stoch hitting 0 for example. Only I don't know, how to define a rule for this.

Sometimes I closed a trade manually by observing with a stoch diff of 20% or more, because the pip result was very ok. Will check trades to the above and will report.

josch
trading is war - life is simple
  • #2,538
  • Feb 10, 2011 4:20am Feb 10, 2011 4:20am
Quoting squalou
thanks for that SMJ, that 's very interesting information...

I think i did something similar in my release, by waiting for each bar open, and i have even added a 5 more seconds delay.

I will try replacing the iStochstich() with a function that calculates it based on iHighest/iLowest() instead, and see if this makes any difference.

If not, then it means it is really the way mt4 reads the history data that is completely f***ed up !!!


Sq
i think the run_once method does work if an indicator uses the 0 bar, but the equity graph only takes the last closed bar and never the actaul bar.

@squalou

i know why your equity looks like this, because you multiply the negativ Lots2 in the code to simulate the opposite direction of trade.

it is a different story if you invert the stoch or invert the traderesult with negativ lot. for explanation in my code:

we compare the two stochs if stoch1 > stoch2 for entry. in EA i coded this would generate trades like this:

OrderSend(FirstPair[x],OP_SELL);
OrderSend(SecondPair[x], OP_BUY);

if stoch1<stoch2 would be

OrderSend(FirstPair[x],OP_BUY);
OrderSend(SecondPair[x], OP_SELL);

and know if you invert one or both stochs it also compare which stoch-value is greater or lesser than stoch2-value and sets the trades.

now what you do with negativ Lots2 in your code:

if stoch1>stoch2 would be:

OrderSend(FirstPair[x],OP_SELL);
OrderSend(SecondPair[x], OP_SELL); //changed direction with -Lots2

if stoch1 < stoch2 would be:

OrderSend(FirstPair[x],OP_BUY);
OrderSend(SecondPair[x], OP_BUY); //changed direction with -Lots2

this is the difference. but with this you do not hedge anymore and risk the double then in previous version.
  • #2,539
  • Feb 10, 2011 4:21am Feb 10, 2011 4:21am
Quoting Mediator
i think the run_once method does work if an indicator uses the...
Use a 1 and not a 0 then.

You are still going to have the problem of the changing history in the off chart pairs unless you lock them down. This is the fault of MT4. You have to lock down the history for the off chart pairs.
  • #2,540
  • Feb 10, 2011 4:31am Feb 10, 2011 4:31am
My test with Trend Filter didnīt open any trade, maybe Iīm doing something wrong, maybe this filter is.....too much filter.

Anyway, today 20 trades with entry at 95%, exit at 5%, triplets and quads, total result (all trades closed): + 65 pips.

I could not compare these results with yesterday since yesterday I used 90%/10%.

Letīs wait.....
8 traders viewing now