Monthly trend of EUR/USD

[ryaron]

I shall give a forecast for the end of May.

I shall be posting monthly forecasts for EUR/USD Exchange rates, monthly.
My analysis is not based on the standard methods used by technical analysts.
It based on phase space reconstruction and projection.
I predict a down trend to circa 1.44-1.45 from the closing value of end of April.

 

[ryaron]

Looks like my monthly trend predictions are converging as they did last month.
No fancy-schmancy indicators and pseudoscience crap.
Just plain old science.

 

[Enivid]

Which science? Can you give us more details?

 

[ryaron]

Science of dynamical systems

Noiseless Reconstruction of phase attractor of time series and projection ahead.

 

[ryaron]

Attractor EUR/USD

The shape of the attractor is quite interesting.

 

[Enivid]

Different colors resemble different time period - blue is the old, red is the new data, right?
But how do you trade using this model?

 

[ryaron]

Different colors refer to different cycles

This model actually shows market "errors".
if today's value is outside one of the trajectories, then it is clear that market will perform "correction".

The trade strategy of such a model is to trade in options.
If the predicted correction is up then a call option is warranted.
And vice versa.

 

[Enivid]

But how do you predict corrections with it? Can you share some link where this model is described in details?

 

[ryaron]

The theory behind it all is called "embedology"

I can refer you to the original papers by Sauer, Yorke and Casdagli.
A good review of Takens theorem is posted here (basis of embedology):
http://eprints.ma.man.ac.uk/175/01/covered/MIMS_ep2006_26.pdf

 

[Enivid]

I can refer you to the original papers by Sauer, Yorke and Casdagli.
A good review of Takens theorem is posted here (basis of embedology):
http://eprints.ma.man.ac.uk/175/01/covered/MIMS_ep2006_26.pdf

But doesn't Takens theorem concerns only deterministic systems? Markets aren't deterministic at all...

 

[ryaron]

Are they stochastic ?
If markets are not deterministic at all, then they are stochastic.
Hence any prediction has a 50% chance to work.

People have proven before that markets show seasonality.
In my opinion the situation is more complex.
We are dealing with a system that obeys a stochastic differential equation (Langevin type).
I.e. the system (market) is deterministic with additional stochastic term.

 

[Enivid]

I am pretty sure markets are dual - there's some dependence between the current and the next state, but there is also an influence of the external factors and the strength of this influence varies greatly in time. That's why some indicator strategies work well during some periods and then suddenly stop working.

 

[Enivid]

Do you use some generic software to build this model or is it something custom-made?

 

[ryaron]

I use

TISEAN to reconstruct the system's phase space.
And yes, you have nailed the spot.
The system is dual.

 

[ryaron]

You are quite right that the short term influences are stochastic, while long term observation shows determinism.
This is quite the approach that I am using.