Trading Strategies Based On Digital Filters - page 54
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Here is comparision to NOXA results on the same signal. They hit 100% on
clean CHIRP signal, 72.2% on _2 and 90.9% on _1.
So definitelly worse even that it is curve fitter type. Can you make it also on GOLD5 ??
KrzysztofSorry but there are not enough bars to have a meaningful simulation.
Sorry but there are not enough bars to have a meaningful simulation.
How many bars does it need ??
Krzysztof
How many bars does it need ?? Krzysztof
I guess 4000 (like the other chirp) is enough but the more, the better.
Anyway, I think that AWGN (which I guess is the noise you mix with the signal) is way too easy to beat in a strategy (as I did).
Can you mix noise distribution of other type?
noise type
I guess 4000 (like the other chirp) is enough but the more, the better.
Anyway, I think that AWGN (which I guess is the noise you mix with the signal) is way too easy to beat in a strategy (as I did).
Can you mix noise distribution of other type?Yes I can mix with other noise, I think intraday market has Poisson noise. I will prepare signal and post.
Krzysztof
GOERTZEL EAs results and final competition results
Here are the results of EAs based on Goertzel V2. I don't have those EAs so I don't know the logic of making buy/sell signals but I know that they are based on cycle finder based on secret
Goertzel V2. Tests were done on the same chirp signal extracted from NOXA charts and I was provided with results only.
so for signal _1 hit 66.7% with PF 5.56 and for signal _2 hit 53.2% with PF 1.43 and 66.07% with PF 2.72.
so it looks that in this competition MESA won, 2nd NOXA and third GOERTZEL as a clear underperformer.
Anyway, now we have an overview how three different spectral analysis methods perform money wise at least for chirp signal with gauss noise the same in all cases.
Krzysztof
to richcap
thanks for you fine work !
for example, I want make spectr on data from 2008.07.01 till 2009.01.01, what set parameters I need ?
sorry for my english..
Yes I can mix with other noise, I think intraday market has Poisson noise. I will prepare signal and post. Krzysztof
Hi Richcap,
Try this one
t = (0:5000)';
f0 = 0.075;
ph0 = pi/6;
f1 = 0.125;
ph1 = -pi/6;
x = 10+0.1*t+2.5 * cos(2*pi*f0*t + ph0) + 3 * cos(2*pi*f1*t + ph1) + 4 * randn(size(t));
It has linear trend, DC component, 2 cyclic components and noise. Later I will replace randn with Poisson noise. Maybe Goertzel 'team' will also make a test on this signal so we be able to compare again performance ??
FFT spectrum is completelly killed by linear trend and DC component.
Krzysztof
Hi dvarrin, you're welcome. I don't have any specific doc, but the code is commented. If you know AT&CF (you can read the two papers at fin-ware site) you can understand how it works. For most of the crosses and timeframes the parameters are quite suited. You only have to find a strategy that fits your way of trading.
So the indicators you have implemented are based on MESA and DFG is based on MESA too? And the number of bars we can choose in DFG is the same like the number of bars you're using in your indicator to perform the spectrum analysis?
About the number of bars to use. Why do you want to take the most recent price history? It looks really noisy with no well defined cycle in it. What I'm doing is to take more and more bars, increasing by 1000 or 2000 bars, until the spectrum analysis lokks almost the same. This would mean that the peaks in that spectrum where significant for a long time and then should not be so bad?
If we look at the adaptive indicator showing the values for P1 and D1, we can see that the curve is quite smooth, exept at some places where the values are making a big jump to another value, before returning to the normal one. Don't you think that the best is to ignore those jumps?
According to cycle analysis, the time spend between two minimums of the price is not decreasing or increasing a lot compared to the previous time spent.
Hi Richcap,
Try this one
t = (0:5000)';
f0 = 0.075;
ph0 = pi/6;
f1 = 0.125;
ph1 = -pi/6;
x = 10+0.1*t+2.5 * cos(2*pi*f0*t + ph0) + 3 * cos(2*pi*f1*t + ph1) + 4 * randn(size(t));
It has linear trend, DC component, 2 cyclic components and noise. Later I will replace randn with Poisson noise. Maybe Goertzel 'team' will also make a test on this signal so we be able to compare again performance ??
FFT spectrum is completelly killed by linear trend and DC component.
KrzysztofKrzysztof,
maybe it would be better to have a slower signal for a meaningful simulation.
As you can see, MESA can extract perfectly the signal peaks, even without denoising or detrending, but it happens that they are too quick for a trading strategy based on cycles.
If you have an 8 bar period (0.125 frequency) or even a 13.3 bars period ( 0.075 frequency), it means that you have 4 (6.5) bars for fast uptrend and 4 (6.5) for fast downtrend. Even a spectacular digital filter introduces some lag, say at least 2 bars .So you have only 2 (4.5) bars to trade a fast trend. Sum it with a noise of amplitude bigger than the signal (4 towards 3) and you have a completely non-cycle-tradable signal.
I like the idea of testing a signal with 2 or 3 sinusoids + DC component+linear trend+noise of different types. I would suggest something like 20bars (0.05 freq), 50 bars (0.02 freq), 100 bars (0.01 freq).
to richcap
thanks for you fine work !
for example, I want make spectr on data from 2008.07.01 till 2009.01.01, what set parameters I need ?
sorry for my english..You have to think in terms of bars. How many bars is your timeseries from 2008.07.01 till 2009.01.01? Obviously it depends on timeframe. For a daily timeframe it should be something like 120-140 bars (which is few). For a H4 it should be something around 480-500, wich is fine.
Put the number of bars in 'length' parameter.