thread: Synthetic hedges, cointegration, mean reversion and similar stuff

Has anybody successfully (with profit) tried to exchange things such as this:



I've opened a demo short on the above synthetic instrument (each point is a M15 candle, the numbers should 1:1 interpret to oanda units if I didn't receive the calculation wrong! ¹ ) to determine how this acts. It seems I can always locate such ranging combinations. The question is just: Can it go through the range an additional time or will it explode now since I've seen it?

______
¹) [edit:] The numbers are really wrong.

This one is interesting:
http://www.google.com/url?sa=trct=jq...AzHq1VWHR5Zv_g

Hi guys,


Basically, I utilize Johansen Method to estimate the co-integrating relationship aka the slope coefficient aka hedge ratio.

The evaluation does not give signs of much co-integrating relationship between EUR GBP, as compared to the version using lm after which the ADF test.

Anyone with the similar approach which we may talk about?

BTW, I bump into a materials saying when there are multiple beta (aka relationship gt; two ), we need to add economic limitation to discover beta. Hence I just stick with two beta


Till now, my back evaluation result still has not guaranteed consistent yield for both (lm and Johansen) method usage. Any 1 kind enuff to share the trading egy?

Hey Guys,

Anyone still trading this ideas?

Something perplexing me for the past few weeks. I have done a Johansen test on most of the currencies pairs, such as EURUSD/USDCHF, I can not find any cointegration relationship whatsoever and the cointegrating relationship estimated by the ca.jo(R function) doesn't work in any way. The residuals not gaussian noise.

I thought the regular procedure is:
1. identify the co-integrated pairs
2. Match the model with lm/ pca / johansen method and
3. Trade the spread

The pairs got to be co integrated to create statistical sense? Else, we're just over fitting some datas?

Yet, 7 pieces, thanks to the effort. I truly do learn alot from your code. Continue the good work and thanks to your willingness to talk about ur code together with us

Really appreaciate it

I try to use arbomat but it do not work, there's not anything desplayed. It use RDEBUG 2 an the path is good. Here is the DebugView



Thanks

I've read through this thread and triggered the arbomat on my computer and am creating a automatic system that tests for significance and cointegration on HTFs to try to find where the optimal entries are across all currencies. Is there anybody out there still listens to this thread?

I have read through this thread and triggered the arbomat on my computer and am creating an automatic system that tests for significance and cointegration on HTFs to try to find where the best entries are across all currencies. Is there anybody out there still subscribing to this thread?

All former contributors woke up one day and found out they had been wiped out.

I have read through this thread and triggered the arbomat in my pc and am creating a automatic system that tests for correlation and cointegration on HTFs to attempt to find where the best entries are across all currencies. Is there anybody out there still subscribing to this thread?

I think cointegration just does not work (anymore) on currencies. The work performed here many others and by 7bit is a fantastic proof of concept and of price that is great and I am sure can be implemented elsewhere.

I guess this is where I am having. No matter -- if anybody is interested, I'll come to my own conclusions and share them. Thanks for chiming in pemully and Ultimate6.

I think cointegration just does not work (anymore) on currencies. The job done here many others and by 7bit is a proof of concept and also of great acedemic value and I'm sure can be implemented everywhere.

Hey Ultimate,

I came to similar conclusions a while back but also feel (as you appear to) that additional analysis of the concept could yield something rewarding. If you or some other developers with R experience want to create a work group to compare scripts and test ideas, please send me an email through this forum.

Cointegration is developed on the idea of regression of a couple of time series in order to discover a long-term terminal combination that may yield stationary residuals which may be acceptable for pairs trading. Is an mechanism for minimizing the sum of the differences between the data or regression points and also a direct line. The very first question which needs to be asked of this method is whether it is sensible to presume that a line will capture the essence of a nonlinear time series such as FX rates. That's pretty much a whopper IMNSHO. However, financial time series do exhibit a property of being.

It's very easy through optimization to discover solutions that look stable but do not continue out of sample stable. The next question that comes to mind is whether it is sensible to presume that by multiplying with a pair or pairs with one coefficient, that the resulting series will be stationary into the future. I believe that is rather a lot to ask of a coefficient or coefficients. The SVECM was made to allow a model to adapt to changing conditions.

The ADF test is a way to determine if the residuals of a regression are stationary. As it says nothing about future stationarity it is a evaluation. From sample testing is a evaluation that is more powerful because it answers the question regarding prospective stationarity of a spread rather than historic stationarity.