Bitcoin Forum

Economy => Speculation => Topic started by: organofcorti on April 12, 2012, 08:50:21 AM



Title: Using ADF or PP test to detect trends in currency markets.
Post by: organofcorti on April 12, 2012, 08:50:21 AM
Anyone successfully used the ADF (http://en.wikipedia.org/wiki/Augmented_Dickey%E2%80%93Fuller_test) or the Philips Peron test (http://en.wikipedia.org/wiki/Phillips%E2%80%93Perron_test) to look for trends in currency markets? It looks promising but also promises a steep learning curve. It's based on a statistical data analysis, whereas many technical trading rules are rules of thumb, or magic.


Title: Re: Using ADF or PP test to detect trends in currency markets.
Post by: ElectricMucus on April 12, 2012, 11:57:53 AM
Well you still would need some model to make some kind of prediction.


Title: Re: Using ADF or PP test to detect trends in currency markets.
Post by: organofcorti on April 12, 2012, 12:39:35 PM
If you can accurately determine whether there's an underlying trend or just random variation, then your prediction is that prices will continue in the direction of the trend. Once the analysis shows the trend is no longer statistically probable, you close your position.


Title: Re: Using ADF or PP test to detect trends in currency markets.
Post by: ElectricMucus on April 12, 2012, 09:58:41 PM
I have to admit I haven't looked into it detailed enough in order to justify a detailed response.
That was just a gut feeling, namely that you would need for example a neural network at some point of going from API data to a time & order price.

I don't know if I have the nerve to look into it, but thanks.
If you know of any further information on the subject I will appreciate it!  :)


Title: Re: Using ADF or PP test to detect trends in currency markets.
Post by: organofcorti on April 13, 2012, 04:17:58 AM
I have to admit I haven't looked into it detailed enough in order to justify a detailed response.
That was just a gut feeling, namely that you would need for example a neural network at some point of going from API data to a time & order price.

I don't know if I have the nerve to look into it, but thanks.
If you know of any further information on the subject I will appreciate it!  :)

Ah, see what you mean I think. I would n't see this being used for HFT, just to give long term traders more certainty about when to trade.