Synthetic hedges, cointegration, mean reversion and similar stuff | Page 4 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Simple Mean Reversion 621 replies

Pairs Trading: Reversion to the Mean 22 replies

Forward test of my new mean reversion strategy 11 replies

Question for synthetic and non-synthetic currency pairs 0 replies

Cointegration, Synthetic hedges, mean reversion in R, Tech Thread 0 replies

Attachments: Synthetic hedges, cointegration, mean reversion and similar stuff
Exit Attachments

Synthetic hedges, cointegration, mean reversion and similar stuff

  • #61
  • Nov 15, 2010 4:28pm Nov 15, 2010 4:28pm
This one seems to behave nicely. (at least today it did )
Attached Image (click to enlarge)
Click to Enlarge Name: 2010-11-15-232422_1023x514_scrot.png Size: 63 KB
  • #62
  • Nov 15, 2010 5:37pm Nov 15, 2010 5:37pm
Hi 7bit
Thank you for your brillant work

Sorry for ask you where can i find rterm.exe.

Sorry again. I do my best to find this exe but i can't success.
Regard
Abroco

PD. Sorry my english. It's no my first language
  • #63
  • Edited 6:16pm Nov 15, 2010 5:59pm | Edited 6:16pm
Quoting abroco
Hi 7bit
Sorry for ask you where can i find rterm.exe.
You need to install R www.r-project.org I thought this was clear from the mt4R thread. Then it will be in a folder that looks like c:/program files/R/<version-number>/bin/Rterm.exe or similar, change the path in the mql4 file to point to your installation of Rterm.exe. Do NOT remove the --no-save at the end.

mt4R.mqh+mt4R.dll is just the interface between MQL4 and R, it contains no mathematical functionality in itself.
  • #64
  • Nov 15, 2010 6:14pm Nov 15, 2010 6:14pm
Be careful betting real money on this approach. It is based on an overfitted model; the linear relationship only holds for the period of time that the regression was trained on, and can be expected to behave differently in the immediate future.
  • #65
  • Nov 15, 2010 8:46pm Nov 15, 2010 8:46pm
Quoting 7bit
You need to install R www.r-project.org I thought this was clear from the mt4R thread. Then it will be in a folder that looks like c:/program files/R/<version-number>/bin/Rterm.exe or similar, change the path in the mql4 file to point to your installation of Rterm.exe. Do NOT remove the --no-save at the end.

mt4R.mqh+mt4R.dll is just the interface between MQL4 and R, it contains no mathematical functionality in itself.
Thanks 7bit.

I have a question on how to use your tool. I could see that the ratios keep changing. So does it mean, we should keep on buying/selling the currency pairs untill we are in profit or once we enter, we dont have to worry about the ratios.
  • #66
  • Edited Nov 16, 2010 12:02am Nov 15, 2010 11:44pm | Edited Nov 16, 2010 12:02am
Can you please upload the Rterm.exe somewhere else. I tried most of mirrors but none of them seems to be fast enough to download the file.

Thanks


Edit: I have got one from US mirror. Thanks.
  • #67
  • Nov 16, 2010 2:02am Nov 16, 2010 2:02am
[quote=7bit;4174508]You need to install R www.r-project.org I thought this was clear from the mt4R thread. Then it will be in a folder that looks like c:/program fil

Thank you very much 7bit

Now is working find

Regard
  • #68
  • Nov 16, 2010 2:10am Nov 16, 2010 2:10am
Hi,

Yes Old Dog...is currently working +800 $ (in demo off course ;-)

One could simple create an EA that will buy the basket on the “-2*stddev” with a stop loss equal a 1/3 of the possible target (which could be the blue line or even the opposite +2*stddev).

About 4320 bars (3 weeks on 5 minutes) .
Buy on condition (-2*stddev) wait for a stop loss or target reached, close if none in 2 days (should reach a condition within 700 bars …) than recalculate the new basket and start the process again.

Ps: 7bit …you provide us with a master piece work! THANKS!!!
Attached Image (click to enlarge)
Click to Enlarge Name: 161120100906.JPG Size: 81 KB
  • #69
  • Nov 16, 2010 4:46am Nov 16, 2010 4:46am
Quoting reachjj
I have a question on how to use your tool. I could see that the ratios keep changing. So does it mean, we should keep on buying/selling the currency pairs untill we are in profit or once we enter, we dont have to worry about the ratios.
I cannot give exact trading instructions yet, I'm still only experimenting with it myself, but I have the feeling that this has some great potential.

It try to find combinations that are stable enough to not change significantly until I am in profit. By changing the size of the time window and observing how the ratios and the curve behave over time I can find combinations that seem to look stable and promising.

I also try to adjust the position sizes if it does not bring me profit after some time and hope that I can at least reduce the loss or break even, eventually it will go my way and give me back some (or even all) pips. I did not yet make an overall loss, it seems when used carefully it is not very difficult to make profit.
  • #70
  • Nov 16, 2010 4:53am Nov 16, 2010 4:53am
superluz, if you increase the setting for base_units from 100 to 1000 (or even more) you will get more precise numbers for the lot sizes. It is a factor to adjust your overall position size and will be multiplied with all ratios before they are rounded.
  • #71
  • Nov 16, 2010 6:07am Nov 16, 2010 6:07am
Quoting realize
Be careful betting real money on this approach. It is based on an overfitted model; the linear relationship only holds for the period of time that the regression was trained on, and can be expected to behave differently in the immediate future.
This is a valid concern.

But if the data contains any sort of signal and not only noise then even an overfitted model will not completely ignore this. And I believe (from comparing models that have been fitted to different time windows) that there indeed is something.

For example no matter how I chose the time window, the coefficients for AUDUSD and USDCAD always tend to have an inverse sign, so it is not like fitting it from two completely independent random walks. The same seems to be true for other pairs. If I were only fitting to random noise then the coefficients would also turn out to be completely random. And this seems not the case.

But I agree that one should not blindly bet his house on these models, it needs more careful examination and some combinations seem to be less random than others. So far I have only visually compared them, I did not make any detailed statistics about how good and how stable the models are. But some of them *look* (and feel) really promising.
  • #72
  • Nov 16, 2010 9:50am Nov 16, 2010 9:50am
did today two trades and now with success. one in range mode and one in trend mode. it seams very simple. another question i have is, if spreadline is at the upper dev. lines do i have to open trades in the direction the lotcoefficients shows, too? or do i have to reverse them?
  • #73
  • Nov 16, 2010 10:01am Nov 16, 2010 10:01am
Quoting Mediator
did today two trades and now with success. one in range mode and one in trend mode. it seams very simple. another question i have is, if spreadline is at the upper dev. lines do i have to open trades in the direction the lotcoefficients shows, too? or do i have to reverse them?
Can you tell us which tf you used this and on what basis you took trades.....which gave you success
  • #74
  • Nov 16, 2010 10:04am Nov 16, 2010 10:04am
ok. i tested for my self. i have to flip the direction. this means is spread is at the upper dev. lines open trades in the reversed sign of lotcoefficients.
uhh. sound too simple to be true. what i'd like to experience is the shifting.

if you look at the video at vimeo you see what i mean. btw nice blender arts.
do work in 3ds max.
  • #75
  • Nov 16, 2010 10:17am Nov 16, 2010 10:17am
Quoting Mediator
if spreadline is at the upper dev. lines do i have to open trades in the direction the lotcoefficients shows, too? or do i have to reverse them?
If you want to short the spread you will have to reverse all signs.
  • #76
  • Nov 16, 2010 10:19am Nov 16, 2010 10:19am
Quoting Mediator
if you look at the video at vimeo you see what i mean. btw nice blender arts.
do work in 3ds max.
No, this was just some amateurish playing around with these tools for fun, I don't do any 3d graphics stuff.
  • #77
  • Nov 16, 2010 10:26am Nov 16, 2010 10:26am
I shorted this (unfortunately *much* too early). But I am still holding it short.

Attached Image (click to enlarge)
Click to Enlarge Name: 2010-11-16-172400_1023x636_scrot.png Size: 57 KB
  • #78
  • Nov 16, 2010 10:30am Nov 16, 2010 10:30am
Quoting 7bit
I shorted this (unfortunately *much* too early). But I am still holding it short.

Attachment
where exactly you shorted? can you please mark it in your chart...
  • #79
  • Nov 16, 2010 10:33am Nov 16, 2010 10:33am
Quoting reachjj
where exactly you shorted? can you please mark it in your chart...
slightly below 1*stddev (the first green line).

But i made positive pips yesterday and closed a profitable long this morning on the same instrument, So I'm not worried yet. Its an experiment. The lot sizes are incredibly small for my account. Lets wait and see what happens.
  • #80
  • Nov 16, 2010 10:41am Nov 16, 2010 10:41am
Quoting 7bit
slightly below 1*stddev (the first green line).

But i made positive pips yesterday and closed a profitable long this morning on the same instrument, So I'm not worried yet. Its an experiment. The lot sizes are incredibly small for my account. Lets wait and see what happens.
I am seeing the below chart for my terminal. Can you tell me how to trade this.. Current stddev is -4.1

Since it is -ve can I buy with those ratios showing in the chart?
Attached Image (click to enlarge)
Click to Enlarge Name: Mine.JPG Size: 61 KB
0 traders viewing now