View Single Post
  #21 (permalink)  
Old 10-08-2007, 04:03 PM
chivas88chi's Avatar
chivas88chi chivas88chi is offline
Junior Member
 
Join Date: Jun 2007
Posts: 26
chivas88chi is on a distinguished road
which one do you like ?

EA is categorized as:

=======



1. EA that use the probability
EA which is
believe that there are good probabilty of
a price breaktru

for example:

pivot EA

or something like this:

stastician have a good xx % probability
that the last high substract by today low
more than yy pips
for the last ten years in a pair

mostly doesn't use "time series forecast"
this EA use formula with variabel of previous data
like last hi, last lo etc


=======



2. EA full indy , that use the indicator as executor
open or close depend on the signal
of the indicator
mostly use trailing stop and not prefer to use TP


=======


3. EA that that use tight rule of TP and SL and
use indicator only to determine the direction of the chart



======



4. EA that has learn ability / artificial inteligence EA, i call it as AI-EA, we will learn about this, after that.. :-)
the input is "output + input"
the output is multi layer weight matrix
process "input pattern" with "multi layer weight matrix = matrix" to produce "output pattern"

input pattern could be combination of indicator pattern, formula and previous price/bar/candle parameter/pattern
output pattern could be amount of reversal, produce about when the reversal happened
in the future within certain period,

this artificial inteligence EA
has an ability to show us when and how many reversal
that will happen during X period in the future

use some genetic algorithm coding
for example backpropagation, scale conjugate, fuzzy, or other "supervised neural network",

this EA choose their own combination of a few indicator and its period/parameter
and do some systematical trial and error to make the best combination
of formula or indy include its parameter that represent more then 80% accuracy of the price movement

the final output of the algoritm is multi matrix
it could be fundamental matrix or technical matrix

to train of this EA need a huge memory and more than ten years data, why?
example: 500 week to the past is needed to train the AI-EA for making 65 week ahead forecasting
:-( i don;t thing that's a cheap stuff
Reply With Quote