andreatrade1 wrote: Sun Dec 30, 2018 1:41 am
Dear All,
should be possible to develop an indicator as per the attached paper?
It's a new approach to the Kalman filter and could be useful for developing a new set of indicators.
Thanks in advance,
Andrea
Normally, an indicator follows the price. The input x_t of the system is price and the output y_t is the result of some equation(s), an mathematical manipulation of numbers. Here, the state space x_(t+1) and y_t (your mathematical model) switches output and input so you have the indicator as input and the price is the predicted output. The prediction error is the difference of predicted value and the real value. How good is your prediction completely depends on your system identification from that you derive the state space. In the paper they took EMA and some distributions of EMA, but they used them incorrect and !as single input! This doesn't make the state space better than the EMA itself. I mean, an chaotic system like FX needs some more filter inputs than a single EMA to be predicted well. Every beginner could tell. This makes sense, i hope.
The result they present in the paper is pretty bad - from the point of a trader. You can save your effort. Take an Hull MA and you have the same result. On this forum are some Kalman Filter indis already you can try.