Non-Linear Cointegration Trading Model

[nononsense]

Quote from man:

my experience is that a good coder will usually prefer to work as low within the environment as possible. and (s)he will do most of time things faster if you look at project as a whole. what you might win with more easy to use dragAndChopAndHopDiHop in the first place, you will often loose when it comes to the ifs and whens in the real world output ... bugs and interpretation and understanding what is really going on ... that kind of thing.

"Les gouts et les couleurs ne se discutent pas."

I believe however in sound engineering in the sense that an optimum way always exists, perhaps unkown to the the practitioner.

"my experience is that a good coder will usually prefer to work as low within the environment as possible": this is simply job security and this costs his boss more money.

About quant-traders loving MathLab: they don't necessarely make money with it.

 

[Equalizer]

Quote from stephencrowley:

I agree with this in general, trying something and realizing it doesn't work does have some value. But a *year*?! It only took me about a day or two playing around with cointegration models in matlab to realize it was well worth it.

Correlation and cointegration are related but very different. Correlation is like, two people walking down the street side by side. cointegration is like, two people walking, one person stops to tie his shoe, runs up, has to wait on the redlight, and then sprints to catch up again.

Stephen, don't give it away man, let them think it doesn't work. Also agree on the time to figure it out, if it takes you more than a few days - at most - to figure this out, then maybe you need to skill up on the Maths/Stats/Programming required before dismissing it. Of course, this does not imply that it will ever work for you. There are guys making serious dosh using a moving average, and guys who couldn't trade using the most sophisticated models even if their livelihood depended on it.

At the end of the day who gives a shit. Use what grooves your equity curve.

 

[man]

dear silentwiser

you are a funny one. no, no, a funny one ...

 

[nononsense]

We seem to have hit a little divergence of opinion on the time it takes to figure something out as being useful. You guys make me nervous about my own figuring out. I'm not going to tell you more about this.

Quote from man:
... to let someone do something for a year without knowing whether it will pay off or not.

Quote from stephencrowley:
I agree with this in general, trying something and realizing it doesn't work does have some value. But a *year*?! It only took me about a day or two playing around ...

Quote from Equalizer:
Stephen, don't give it away man, let them think it doesn't work. Also agree on the time to figure it out, if it takes you more than a few days - at most - to figure this out,...

Quote from Equalizer:
...
At the end of the day who gives a shit. Use what grooves your equity curve.

I bet, that's the way we all like it.

 

[man]

the thing that takes time is not the single whateverAlgorithm that replaces something in bigger context be it cointegration, wavelet and so forth. what takes time is move somewhere totally different. we were only used to single time series five years ago. then turned to real market neutral testing including cleaning of database from biases, bla bla. once you have the frame ... well ...

i am still reluctant to go into level II data. still think uncorrelated sharpe can be obtained easier, quicker elsewhere.

i have something lying around for japanese equity pairs trading. the testing took a week, but i know it will take me months until everything is really, really in place to switch on trading.