i have a question to fxtimer,crodzilla,codebreaker and other experts...instead of simply adding up all the waves, is there a better way to find which cycles you want to use in the current bar?thanks,deivi
Optimized Risk vs Reward Equation 335 replies
Loss Less Optimized Strategy 63 replies
Optimized Parameters in Metatrader Testing? 3 replies
"Optimized Parameter" feature in EA? 0 replies
DislikedIf this is something you are interested in, you can find it on another forum; I'll send it to you if you need it; I can't post it because it is not in the public domain.Ignored
function y=asm8086filter2(open,high,low,close) %y=asm8086filter2(open,high,low,close) % %asm8086 Filter from post #699 %http://www.forexfactory.com/showpost.php?p=3371056&postcount=699 N=length(open); if(length(high)==N && length(low)==N && length(close)==N) if(N > 116) [hopen,hhigh,hlow,hclose]=HeikinAshi(open,high,low,close); %#ok<NASGU> x=(hhigh+hlow).*0.5; H0=ullma(0); H1=ullma(1); h0x=H0.filter(x); h1h0xd=H1.filter([0;diff(h0x)]); H0.States=0; h0h1h0xd=H0.filter(h1h0xd); y=h0x+(h1h0xd.*2-h0h1h0xd); y(1:116)=0; else error('input vectors length must be greater than 116'); end else error('input vectors must have the same length'); end end function [hopen,hhigh,hlow,hclose]=HeikinAshi(open,high,low,close) %[hopen,hhigh,hlow,hclose]=HeikinAshi(open,high,low,close) % %Heikin-Ashi Indicator N=length(open); hopen=zeros(N,1); hhigh=zeros(N,1); hlow=zeros(N,1); hclose=zeros(N,1); hopen(1)=open(1); hhigh(1)=high(1); hlow(1)=low(1); hclose(1)=close(1); for n=2:N hclose(n)=(open(n)+close(n)+high(n)+low(n))*0.25; hopen(n)=(hopen(n-1)+hclose(n-1))*0.5; hhigh(n)=max([high(n);hopen(n);hclose(n)]); hlow(n)=min([low(n);hopen(n);hclose(n)]); end end function H=ullma(type) %H=ullma(type) % %Codebreaker's "Ultra Low Lag Moving Average" Clone %http://www.forexfactory.com/showthread.php?t=68181&page=10 Length=39; switch(type) case 0 Cycles=1.25; case 1 Cycles=2.5; case 2 Cycles=3.75; case 3 Cycles=6.25; otherwise error('Unknown type'); end x=(0:1/(Length-1):1)'.*(Cycles*pi); h=sinc(x); h=h./sum(h); H=dfilt.dffir(h); end
DislikedIm not sure about this Heikin-Ashi candlestick plot.
Your candle stick is already a filtered version of a normal candlestick, so your graph isn't showing the true price.
Hence your trades are unlikely to be as clean as thatIgnored
DislikedAlberto,
did you use this tool for your forecast ?
http://www.advancedsourcecode.com/lpctrader.asp
SixerIgnored
DislikedGood morning you all guys,
very interesting post and a lot of great math and dsp skill people here!
Someone before wrote about Bartels, Fourier and so on; following Herbst's book I have developed the whole process in Matlab, with some modifies.
Results are very good but I wish to improve and I wonder if there is someone interested in collaboration. Dsp and Matlab knowledge is required.
Than you very much....Ignored
DislikedHello Sixer,
as I stated previously, all job was done using Herbst and Bartels approach as described in "Analyzing and forecasting future prices", but with some modifications to the original concepts.
I don't even know what this link is about.
I have nothing to sell, only I'd like to improve what's done so far.
Regards,
AlbertoIgnored
DislikedLast year you were trying to find some people to work with the same and showing the same picture. So what was your results during this year ?
Are you able to post in sample and out of sample real trading results ??
What is your method to distinguish that this picture is a result of real predictive power not just luck ??
Herbs't book is quite old. Why you assume that what he writes is correct ?
KrzysztofIgnored
Consultant on the problems of financial markets forecast, The Market
Forecast Group, Italy
Tasks
: Development of an approach to the automatic forecast for Forex
and Italian financial markets based on the Caterpillar-SSA method,
DislikedMay u could be interested also in SSA for forecasting, here is a stan alone free program and also the code for Matlab http://www.pdmi.ras.ru/~theo/autossa/english/index.htm.
Lubo.Ignored
QuoteDislikedSecond question: no I am not able to show any in sample and out of sample result since this is not build as a trading system.
Maybe it is possible to turn as a TS but this is not my goal
DislikedAttachment
Here is something pretty close to CB's clutter filter. I program only slowly in MT4, but do okay in afl, so I did it on AmiBroker. This is the idea:
Use the ULLMA of HA as suggested by ASM8086
Let Noise = HaHigh(ULLMA(OHLC)) - HALow(ULLMA(OHLC))
Let Signal = K*slope(ULLMA((HaHigh(ULLMA(OHLC)+HaLow(ULLMA(OHLC))/2)
Let SNR = 20 log(Signal/Noise)
Plot yellow bars if SNR<Threshold. (This is clutter)
I do not have the ULLMA in AFL, and did not want to write it there, so I used Igorads NLMA with length 15 as a proxy....Ignored