Digital ACSTrend - page 34
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
Please be aware that those are only instruments and you are fully responsible for your actions.
If we make general statements, this or that does not work that leads to nowhere. Please let all be specific what exactly did not work.
The testing is something very specific issue, thanks for the information and the links.
Here are the results from 2007 with length 70.
.......
Hi John Last,
Backtesting results can prove nothing (it is proven that nothing - by many threads and posts). It is experience of our forum since 2005. Especially if EA or system was coded/used in MTF way, open bar trading, high/low of the close bar, volume-like indicators, jurik indicators and so on.
Usually we are banning the members who are trying to sell any EA based on backtesting results (especially if we do not know how it was coded by source code).
We can use backtesting to create the settings for EA (in some cases, not for all EAs, and not for MTF EAs).
I traded the systems. For example: ASCTrend system. Traded and traded many times to repeat and repeat.
And it is my latest from advanced elite - I traded Helweg / Stendahl value charts indicators 3 times by different setups.
You can see:
And in zip file too:
EURUSD: 457 pips by 43 trades;
GBPUSD: 3 pips by 2 trades;
GBPJPY: 195 pips by 53 trades;
GBPCHF: 268 pips by 52 trades.
So, it was 150 trades (opened and closed manually with manually placing stop loss and take profit) with 923 pips (4 digit pips).
MANUALLY! Every single trade was opened manually with stop loss and take profit, and closed manually.
Not backtesting.
Forward testing (means: trading).
===============
To make it shorter - no any conclusion about profitable/non-profitable can be made based on backtesting results.
Your thread is very good and popular for now: many people sent PM to me about it telling that they like what you are doing. So, please, do not ignore my suggestion and believe in my experience (and in experience of the members of the forum): trade your setups for at least few days and it will be much more valuable that any useless backtesting results.
or ask someone to trade ...
Create 1 or few setups of indicators which can be ready for trading (manually), without MTFcases and without repainting/re-calculation issues - and people will trade and will post the results if positive.
Hi New digital, Hi all,
This is is not my thread this is our thread
.
First nice results.
And thank you again for the comments. Well my intention is not to sell anything.
To Francis I would say I that we have discussed SSA ep and its use in EA. And the conclusion is, It is difficult, you cannot make it, I cannot make it and some other guys tried and did not make it. Thank you for the collaboration.
I wanted something light and it is the double smoothed jurik: fast and reliable
The main reason for making those tests was to know historically what is going on. That is because the results look too good for me. And the results showed that applied with the simplest strategy it is profitable and right now it works (not like a money machine).
Second I wanted to know what time frame is the best to trade.
To tell the truth this project was together with a professional trader who is a little bit as a mentor for me. He told me that the results are not OK for him because he looks for much bigger profitability. But the question is what is OK for me.
And it is OK for me because I am looking for a global solution. This system offers me that. The main point is that I change only 1 parameter (length) and at almost all values the results are positive by now.
Second and this is my theory. I am looking for another system that will guide me locally. And I do a local optimization for the current market conditions.
Basically I used Neuroshell with wavelets, because that is easy.
But with an expert we can export data to Rapidminer and have a trading solution there. And we have a myriad of good indicators here on this forum that merit to be used. Basically I was thinking about A-Trous wavelets with Support Vector Machines of RapidMiner (a local solution).
For me the bigger question is to know when to optimize and for how long.
Some posts ago I used a classification by Curtis Faith together with the fractal dimension. It is a possible solution. For example we have a trend with low volatility. We optimize on this, from the start of this market condition till today.
https://www.mql5.com/en/forum/180729/page8
However we have a principle of uncertainty. The more the market condition has developed (we have a perfectly optimized local solution) the more it is probable to change.
On the contrary when the market condition has just started. For example we had a high fractal dimension with range and we had just a break - out with a fractal break - out. In that case the Neural Nets that were trained for the market conditions before start to fail one after another.
In that case once the market stabilizes a bit we start to train again. Well we have a few data but with this data we have a more robust locally optimized solution.
That is the idea. I had even a suggestion for an innovative way to use the Elliott wave principle because it uses an advanced form of pattern classification. For example you identify a complex correcting pattern and you train and validate within the pattern. Even if the count is not right it does not matter, what matters is to identify the beginning and when it will end. And it ends with a fractal break - out.
For MT4 for local optimization can be used an ASCTrend expert locally optimized. In this case the dots are globally optimized and the arrows are locally optimized for the current market conditions.
Roughly speaking those are the ideas, far, far away from setting a robot with back-testing results making money for you.
Look here at this shot. This is just before the tremendous down turn of the Euro. We are just at the top and we had a range. I used several different inputs with different architectures. I have identified a local condition low volatility range and I trained on that. All the nets gave consistent results the green bars show forward testing testing results as the chart was saved. And after the break - out. Then the nets started to counter the impulse. And that can be understand it well, they were not trained on that. they do not know that stops were gunned one after another.
Just remark the little optimization and validation period.
SSA indicators and reliability
Well
I would like to add something regarding SSA indicators
As far as I can see, it seemed SSA indicator allocates memory internally, so, it may fail the second time the EA calls and worse, I had several MT4 demo accounts on # brokers( GCI, MBT ,FinFX... w/ SSA and Nikitin library installed.) for testing. I dunno why but FinFX failed. I did re-installation, but always same kind of failure.
(The GCI and MBT metatrader stations were still working but I can not certify the result from SSA indicators, even SSA indicators are very good to identify trend in the past, this fact does not imply that futur will be predictable using SSA, there is a total disconnection between past and futur w/ SSA indicators. A contrario, divergence can be detected easily w/ Stochastic. But you can not agree w/ what I am saying, of course)
I try to perform direct functions like this one :
#import "SSA.ex4"
void fastsingular(double a[],int n,int lag,int s,double& b[]);
#import
......................................
fastsingular(arryTimeSeries,N,Lag,NumComps,SSA);
As a trader told me, it's because of the way that MT4 calls a DLL. It will then fail at every call. May be, writing SSA libraries for EA's that allocate memory internally, could solve this issue.
IMHO, on Windows environment, MT4 workstations are limited and built for ordinary indicators (ATR, ADX, RSI, OBV, Sto...) , not for SSA indicators, involving plenty of memory access. (may be the solution is to have a MT4 on Linux, I am not sure if MT4s are written for Linux environment and special Unix calls, or if MT4s are just emulated)
Furthermore, I think a better datafeed and a new type MT4-client-station should be provided, for this new kind of SSA indicators. If someone could answer to those subjects :
- where to fetch a better datafeed ? (even there are no dedicated link for MT4)
- which new client station should be used for SSA indicators (even if MQ4 should be replaced by C++) ?
Rgds
john yves
You know what I am doing in this situation?
Just my experience:
Create few templates/systems/setups with full description ( MTF or not MTF indicators, repainting/recalculating etc). And propose those templates/systems/setups for manual trading. Someone (you, I, or anyone who wants) will trade.
Results can be analyzed (professional traders are analyzing losses more than profit).
In this case - the members will have something to try/to test etc.
And you will have some ideas from them about how to improve and so on.
As to repainting indicators... I used them for trading: those indicators are giving some "warning message" about something as "possible sell will be soon" or "possible buy will be soon". Just to be near PC at this time
After that - all you/we need to do is to estimate/predict market condition, and review/re-evaluate those templates telling that ... "template/system #1 works in ranging market on M30 timeframe", "template/system #2 ..." and so on. And after that - every template will have the manual/howto instruction etc.
After that - indicators will be improved and simplified to be used for EA's coding (indicators should be recoded). After that - we will have few good EAs.
Because I am sure - if all those steps will be done so the coders will stand in a queue to help you for coding (similar cases was for regular asctrend thread, for Brainwashing EAs, for mandarine EAs and so on where many coders decided to help in the same time). Because every coder will be happy to code profitable EA.
And after that only - you can backtest just to find good settings for those EAs.
It is what I would do if I would come to similar research.
Just the ideas.
By the way - it is your thread so you can decide anything here
interesting thoughts.
my ea that i made which doesnt do what i programmed it to do is still trading positive.
go figure.
Conversion
Newdigital,
can you assist in solving this?
Optimized Trend Trading - Page 76 @ Forex Factory
Hi Camaron,
We should ask Mladen as he can do it.
But I am not sure ... may be - this indicator is already existing for MT4.
camaron
I am copying the post that was originally posted in the elite section. Among other things we are trying to show if some indicator is what it seems to be or it is not that (a nice mirage to put on selling pages, for example). I am afraid that the indicator in question is the later case ...
Here is the original post (since I think it will explain quite well all that needs to be known about it) :
___________________________
...
Just one thing : but (the obligatory "but"
) be careful how you use it.
Values are not going to change (hence I decided to add the lines and make them more visible) but the signals are going to change. They are searched for as minimums and maximums and it means that there is no guarantee at all that they will stay where the appear for the first time (from the calculation itself it can be seen that signals can change up to window bars back, which means, all the visible bars this indicator is drawing)
See this example : it is moving the signal all the time, and even when after the signal bars of opposite color appear they can be changed whenever a trend continuation causes new highest value (see second example - if EURUSD rises more, those bars at the end are going to disappear) The same thing is going to happen on metatrader 5 so ... just carefulPS: the appearance is a bit different simply because metatrader 4 can not draw non-zero based histogram, so this one needed to be modified to "dance" around zero values PPS: I just noticed that the original indicator has shifted values (it looks better in hindsight but it does not change what has been told about it, it does not change when the signal is given (in real time it will appear at a time of the bar shown on the upper (non-shifted) indicator)) so added a choice whether you want shifted or non-shifted values. Also, some minor code changes have been doneregardsMladen
camaron
I am copying the post that was originally posted in the elite section. Among other things we are trying to show if some indicator is what it seems to be or it is not that (a nice mirage to put on selling pages, for example). I am afraid that the indicator in question is the later case ...
Here is the original post (since I think it will explain quite well all that needs to be known about it) :
___________________________
...
Just one thing : but (the obligatory "but"
) be careful how you use it.
Values are not going to change (hence I decided to add the lines and make them more visible) but the signals are going to change. They are searched for as minimums and maximums and it means that there is no guarantee at all that they will stay where the appear for the first time (from the calculation itself it can be seen that signals can change up to window bars back, which means, all the visible bars this indicator is drawing)
See this example : it is moving the signal all the time, and even when after the signal bars of opposite color appear they can be changed whenever a trend continuation causes new highest value (see second example - if EURUSD rises more, those bars at the end are going to disappear) The same thing is going to happen on metatrader 5 so ... just carefulPS: the appearance is a bit different simply because metatrader 4 can not draw non-zero based histogram, so this one needed to be modified to "dance" around zero values PPS: I just noticed that the original indicator has shifted values (it looks better in hindsight but it does not change what has been told about it, it does not change when the signal is given (in real time it will appear at a time of the bar shown on the upper (non-shifted) indicator)) so added a choice whether you want shifted or non-shifted values. Also, some minor code changes have been doneregards MladenNoted. Thanx for clarifying. Saves time as well
SSA code instead of DLL
SSA as a block transform may be used as a filter. In this format it repaints back about 10-15 bars depending on the parameters chosen. If the end point is captured and plotted, it will not repaint, but it will not show the same values as the repainted SSA. IMHO, the end point can be closely approximated by the T3 filter ( and probably by Jurik of some type). It is not clear that the end pointed SSA is any better than T3. It may be slightly smoother for the same lag, but that's not clear.
Here is an MQH library that implements fastsingular in MQ4 code. It can be used instead of the DLL, and should work in an EA if you really want to try.