- Search Forex Factory
- 27 Results
- fajst_k replied Sep 27, 2015
Well I put the effort. My system based on ML trades 8 instruments on 1 min charts. From attache screenschoots u can see equity curve from demo account from 15.09 for total account and separate symbols. (around 10 000 trades done for 2 weeks). The ...
- fajst_k replied Dec 12, 2011
equity_V7 for backtest ?? — Does it work for backtest ?? i tried and no success
- fajst_k replied Nov 9, 2010
LagLasso and Kalman — And this one claims to predict S&P using LagLasso regression and Kalman filter url easy to check
- fajst_k replied Feb 23, 2010
for sin + cos 600 bars — correct sign 100% most valuable inputs are momentum and lag values
- fajst_k replied Feb 23, 2010
for 576 bars EURUSD — correct sign 62% inputs chosen equally 1,4,5,6 so EMA, HP, HP 2 pole and Butter
- fajst_k replied Feb 23, 2010
for 7k bars EURUSD — correct sign 52.58 most valuable input CLOSE and lag(1)
- fajst_k replied Feb 23, 2010
for 65k bars EURUSD — correct sign 51.6% the most valuable input seems to be HP filter, momentum(1), bandpass and bandstop
- fajst_k replied Feb 23, 2010
experiment — I made some in my opinion interesting test. I wanted to verify if DSP filters are giving valuable input for the strategies when input data series are financial DS. For this i used NS and put outputs from filters as a input to NN ...
- fajst_k replied Dec 15, 2009
you will love it !! — I love this !!! url
- fajst_k replied Dec 11, 2009
Hurst exponent and perfect TF — Here is Hurst exponent calculated for 1min and 4h TF for different lengths. It is calculated with 'Servicik" method which I think is one of the most accurate. Referencial Random Walk 0.5 in blue. It's clear from ...
- fajst_k replied Dec 10, 2009
Thats not homodyne discriminator but another method of cycle measurement based on Hilbert transform. Homodyne was descibed in Rocket Science....lag 20.5 bars...anyway both methods in my opinion dont work well.. Krzysztof
- fajst_k replied Dec 10, 2009
homodyne discriminator !!! — I believe you are not so much updated. John gave up with frequency discriminators in 2007 (Hilbert transform based) see screenchoot. Homodyne discriminator has a lag around 20 bars !!! Just make a simple strategy ...
- fajst_k replied Dec 10, 2009
I think they are simply because of this basic rule. Or put it in another way. If you take let's say 4H TF than to achieve the same measurement error like in 1M TF you need 240 times longer time. Than you will trade much less frequent and make less ...
- fajst_k replied Dec 9, 2009
Perfect TF ?? — Here is a screenshoot of Hurst exponent in relation to length of TF for German Euro Bund. According to this only very short TF are anti-persistent. I wonder how it looks for currency pairs, I think script for MT4 or MATLAB ...
- fajst_k replied Nov 6, 2009
DSS + trendiness — You are welcome. Than I made another test DSS + Trendiness indicator. I smooth original one with 4 bar WMA as momentum function used in indi makes it's very 'noisy'. See screenschoots. So overall DSS PF 0,72 %profitable ...
- fajst_k replied Nov 6, 2009
chopinnes idx performance — Here you have quick test of chopiness index performance. I applied Ehler's DSS strategy and DSS + chopines index to the same data set to 5 currency pairs. Overall it improved yearly return from -153% to - 77.6% and ...
- fajst_k replied Nov 5, 2009
clutter — Are you sure that what CB said here about trading systems is 100% correct ?? Because me not...See for example this - one of the J. Ehler eminiz strategies 8. Dual Super Smoother (DSS) Strategy DSS is an acronym for a Dual Super ...
- fajst_k replied May 7, 2009
lag reason — The reason of such big lag is that the 'search area' for a cycles is 3xMAXPER so 3x observation window. I think it is mandatory condition for Goertzel. Than the spectrum amplitude is calculated for every bar and as long as the ...
- fajst_k replied May 6, 2009
Goertzel lag — Here is the screenshoot for test signal to find out the lag of Goertzel cycles and SNR compared to lag of NOXA CSSA cycles. In both cases observation window (m-history for CSSA) was set to 120 bars. Lag for CSSA seems to be size ...
- fajst_k replied Apr 29, 2009
SNR — Hi, Meanwhile I finished to code SNR for NS and here are the screen shoots This SNR is calculated that signal is a dominant cycle and noise the rest. SNR is calculated based on peaks of spectrum amplitude as a 20log(S/N) so peak for ...