It depends on Cycle Period and the JJMAseries (in the include folder) from the Kositsin library to work.
I will re-check when the markets open,
I like your ideas and we both share some things in common in terms of ideas and creativity. I realy realy fancy your style keep it up.
Picture of the day
The picture of the day. Really looks nice. This is the Mama mod with adaptive JStochastic signal mod.
I am thinking about a filter for this system. For the Advanced Elite members the answer is obvious:
Take a trade only in direction of the slope
For Elite members my answer is:
For the ordinary members I am not sure yet. I really want to opt for a TSD indicator (#_FullSSAnormalize, The Trader's dynamic index and the BB_Macd are nice but they are not TSD indicators).
I made my own mod using Caterpillar which is a free version of SSA of price (I took the original Caterpillar indicator and I have cleaned it up deleting the second line and allowing the change of parameters (lag bars, number of computations and bars for calculation), the original uses 7 and 5 lag bars, I use much more). However the SSA of price is much more stable when you use more lag bars.
Well this is very special indicator, because of the nature of the SSA. It recalculates. But if you apply a sufficient amount of lag (at least 30) it can be very accurate.
This mod is a full SSA mod. We apply the caterpillar in two different ways in order to analyse the trend.
You need the caterpillar and you need to install the library file.
I send you two versions one is bb_squeeze_caterpillar: I use the original idea of the squeeze as identification of trend (the dots). The histogram uses the Caterpillar (SSA)
The second the bb_squeeze_caterpillar_2 uses the SSA (Caterpillar) but in two different ways: histogram and dots. The dots recalculate more often that the histogram.
Another trend filter
Here I add a version of #_FullSSA but here we do not have a normalization.
As for input of the price series I use a moving average if I want to smooth them.
I use for default a period (Periodma) of 2 but you can play with that of course.
In fact the normalization is a deviation (and maybe innovation) from the logic of the SSA. The SSA decompose the price time series.
Singular Spectrum Analysis (SSA) is a method that takes all the variability in a series and breaks it into a few oscillation patterns that we refer to as eigenvectors. It also gives a measure of significance of these patterns or eigenvalues.
“Eigen” is a German word, which roughly translates to “characteristic”. Eigenvectors are structure functions that best represent the modes of behavior in price. The eigenvalues are a measure of the variance that these modes account for.
In the SSA_Normalize we use the component that accounts for most of the variability in the data; it is the lower frequency component.
After that we normalize it by a period of normalization by a moving average.
This is an innovation because we use a scientific instrument and we apply ways of use from the technical analysis indicators. That is not a use a scientific guy would do.
The classical approach is to expore the Singular Spectrum.
You can read the tutorial of CSSA Noxa for 101 of SSA.
Forex Factory - View Single Post - Optimized Trend Trading
fajst _k has a clear position
My opinion is that if you use a sufficient amount of lag you are fine. Of course you need to be modest in your expectations and use as additional trend filter. Of course this tool can never ever predict a break - out. And it has to be used when we have a stationarity in the weak sense, forget about it when we have a fractal break out. The next step is to use a Neural net as price input. I did it but it is under experimental stage.
However with this mod we do not do that we just use the SSA without normalization by varying the number of computations we can explore the Singular Spectrum.
For exapmle if I make 1 computation I will see the eigenvetor that explains the most if the variability of the trend. If I use 2 I will explore the next eigenvector. Here when I go to 5 I had the eigenvector that explains the oscillatory component of the price.
Yes it is a classic but we did not have it under Metatrader.
Well it recalculates but if you apply a sufficient amount of lag it would be manageable.
A possible use is together with trader dynamic index of SSA normalize. The SSA normalize of Trader's dynamic index is giving by normalization the oscillatory patterns (we use only the most important eigenvectors and we normalize them, scientifically crazy but, damn it, it works) and the FullSSA is giving us the trend.
This is a conceptual mod and not a trading system.
/You need the same library/
The original is a normalized version you can find here
Here I post a picture of the Adaptive ASCTrend.
The settings are Risk 1 (Basically I keep those settings, and let the algorithm adapt, if there are too many signals it is necessary to use digital smoothing but it concerns the lower time frames and that is not the case here).
We use Niquist parameter at 0.5
For the mama mod we use
Fast limit 0.6
Slow limit 0.05
Below we have a BBsqueeze caterpillar mod.
This is usefull for two reasons.
1. When we have a blue dot that means that we do not have enough juice for a trend
2. When it is green the Slope should confirm the direction by the ASCTrend system.
I add another shot at that shot the Nyquist parameter is set to 1. That means that we adapt to the length of the dominant cycle. If it is set to 0.5 we adapt to a half a dominant cycle.
So we have four absolutely independent algorithms.
The signal ASCTrend relies on:
jurik smoothing, adaptation to the cycle period, takes into account the volatility and tries to give us signals when we have strong momentum into given direction
The mama mod. Relies on the Mama of Ehlers to go into the right direction. We can optimize by a genetic algorithm this component to have a statistical edge.
The BBsqueeze combines two different ideas. One is the original idea of the squeeze to select when it is good to trade and when not, that is done by comparison of the Bollinger and Keltner channels. We need that the Bollingers are wider that the Keltner channels. This is purely heuristic approach but still is of some use.
And for the Slope is used a Caterpillar (Singular Spectrum Analysis) to determine the direction of the movement.
Well this looks not different from any other technical indicators (and the Normal ASCtrend performs nicely at the current market conditions but it is when the market is difficult that will be the time to prove its worth not now) but the algorithms behind are quite complex and it requires a powerful machine to run that toy.
Any ideas for improvement are welcome.
Here I will not look at the fractal patterns of the FGDI, which is the spirit of the System.
The Stochastic mod is not bad either.
It is more reactive the WPR mod.
Below is my Trader's dynamic index on SSA. Well with this mod you can see clearly where is the trend and if we have a trend or not. This instrument shows us the current volatility. If the volatility is reduced we should not use the same money management as in the trending market. Once a winner always a winner.
And of course if the bands are really close that is not a good moment to take a trade. Wait for a volatility break - out then. The system has never been designed to pick tops and bottoms, well occasionally it does but it is another story.
During a trending market we can use the stops and let the position breath but never in ranging market.
All ideas are welcome. Especially concerning the money management my weakest part.
The only problem i have with the SSA family is their Repainting tendency
Other tools, but the choice is yours
It is not repainting it is recalculating. However if you are not comfortable with that do not use it. I find it useful with enough amount of lag they miss occasionally to recalculate. However I am interested mainly if it is above or below 0 line.
I think that the system is based on the signals and the stop line. Anything else is accessory and modular, and is a question of choice.
Here I add another idea in the tool box:
This is a retraining neural net with adaptive training period of 1.5 of dominant cycle period. I add the original indicator also, more information you can find here (I use the fixed library not the original one).
There is another useful tool that can be combined with this system for better timing of entries in the direction of the trend. (Of course it is constantly recalculating and it is for the Repainting indicator's fans collection)
I use a Neural Net, however it is really a short term prediction . I use it and the training period is constantly tuned to 1.5 of dominant cycle period. It is even more accurate than the default 300 period. The idea is that the market is loosing constantly the memory and I believe it is wrong that if you train on 1 year data you will get better results (there are guys who will kill to protect their believes of course). Well this is my opinion that by tuning constantly to1.5 of cycle period I can get good practical results.
It is a nice toy, play with it. Sometimes the predictions it gives are really close to what a human technician would do. Interesting!
It is mainly used when you want to play into the trend but you do not want to have a stop loss close to the stop loss dots. So you can use this for timing the entry for not being against the odds. As usual its main use is in relatively calm market conditions when the market is stable enough.
Do not use when the volatility is too high.
Adaptive Trend channel
I have always been lazy to draw trend lines.
So I did something. I want to have around automatic assistant who will draw channels for me.
I used an indicator for linear and polynomial regression and I tuned it to 1.5 of dominant cycle. Of course this is another unusual use of the idea, but I like it.
The channel is adapting itself to the market conditions. Please do not forget the Runge phenomenon when using high order of polynomials, I use linear regression mainly.
I post the original and my mod, which is not polished but works for me.