a trading strategy based on Elliott Wave Theory - page 290
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
Briefly about the model: I have realized that for reliable calculation of future movement we need only one channel, Hurst index and "fractal" calculation of movement level. The code takes about 100 lines, it is very simple.
I tested it on Alpari on minutes (at Alpari). On hours I have some glitches with history (I complained above). In general, the algorithm should work with any timeframe.
The only one losing deal - even that was not enough history. The lot is stable 0.1. It trades little, but no errors. It may increase the number of deals, but in this case the risk increases as well. This result is a kind of "maximal forecast reliability". I don't really understand why the quality of modelling is na, maybe because of the minutiae?
And here's the report itself:
http://grasn.narod.ru/TEST/StrategyTester.htm
to Rosh
Thank you, I'll keep digging. By the way, maybe because of proxy the downloading of history from MetaQuotes server does not work - it just hangs and that's all.
Impressive result!
The quality of modelling depends on the model you choose in the Tester:
1) By opening prices ( as you have) ;
2) Reference points ;
3) All ticks
Just one question, the 15th trade from 30.08.2004 to 24.06.2005.
Over sat 16 figures :(
Thanks, I will look for more information. By the way, maybe because of the proxy the downloading of history from MetaQuotes server does not work - it just hangs and that's all.
Yes, in the case of proxy you need to enter the address of proxy server.
You can use DemoCharge2005 v1.1.1.9 compiling GIF file or SnagIt v8.2.3.
And the hoster, for example, http://rapidshare.com
Thanks for the tip on the hoster. Put a video file there (one so far). Those who want it can get it here: http://rapidshare.com/files/42092574/CWT_movie1.Zip.html
Didn't get the rest, but thanks just in case. I do not use any intermediate image files. The result of calculation (after proper scaling) is written directly to the video file via codec frame by frame. No problems - all automatically. Codec is simple and widespread (Microsoft Video 1 - MSVC), so it should play everywhere.
The quality of the simulation depends on the model you choose in the Tester:
1) By opening prices ( as you have) ;
2) Reference points ;
3) All ticks
Ah got it, now I'll run with "all ticks" option. Damn, I forgot everything...
Just one question, the 15th trade from 30.08.2004 to 24.06.2005.
Outlasted 16 figures :(
All right, there's one subtlety I'm wondering how to deal with. Used a very long channel length - 3 months, in this case, a constant value, in the future I'll transfer the code from matCAD for optimal channel length selection. But still. The channel length must be "statistically" large, otherwise the Hurst index, which is highly dependent on the sample length and what is inside the sample will produce biased values.
Next, a prediction of the channel lifetime is calculated using a formula based on statistics (I found a good book on "deriving" formulas). This calculation can give an existence time for a particular channel from 0.1 to 7.0 of its length (roughly speaking, the channel will live another 7 of its length, for example).
It is assumed (if the channel meets the criterion) that the price however it wanders 'necessarily' will pass the calculated level during the 'lifetime' of the channel (in a sense, I proved it scientifically to myself after, I think, 6 beers), and the waiting time can be, as you see, long.
In other words, there is no analysis yet of the "reasonableness" of opening a position if the waiting time is too long. Thinking about how to deal with this, it's not so trivial here.
There is a second way - "directive" reduction of the calculated price level. :o(
And most importantly! Only Hearst works, mostly. :o)
to Rosh
Yes, in the case of proxy you have to enter the address of your proxy server.
Well, it has been introduced, otherwise it wouldn't work at all. But the quotes are coming in but the history is not loaded :o(
It's still not clear. The 14th trade lasted 20 minutes, the 15th tortured for 10 months. Why in the 15th trade the channel showed SELL and not BUY?
On the other hand, if the mathematics is successful even after such a long wait, it should be applauded. But do you have enough faith in the method even at 500 points of losses, and it was 1600? And all that time...
Hi Sergey !
This test unfortunately cannot be considered indicative. In the absence of stops this ratio of profitable to losing trades is in the order of the day. But the lack of stops is a mistake of beginners that is constantly criticized on this forum and makes the results totally biased.
Try to run the same test in "all ticks" mode with stops. Try to optimize the SL parameter. If you get a positive profit of the strategy with SL<MO, this result will already mean something.
Другими словами, пока нет анализа «разумности» открытия позиции, если время ожидания слишком велико. Думаю, как с этим бороться, тут не так все тривиально.
It's still not clear. The 14th trade lasted 20 minutes, the 15th tortured for 10 months. Why did the channel show SELL and not BUY on the 15th trade?
On the other hand, if the mathematics is getting its way even through such overhosting, then kudos to it. But do you have enough faith in the method even when losing 500 pips, which was 1600? And so much time...
So I didn't write that I invented the Grail and I'm not suggesting that I'll give it away for 100 coons. This is the first test in MT model in its purest form (one of the three invented and the simplest). In MathCAD it takes three lines (the kernel itself), in MT it takes a hundred lines. Moreover, I haven't even started to implement fully in MT order management and other service functions as it should be done - I know very well that it's too early.
Of course, this version in its pure form is not working yet. But it is for now. And I'm not really going to sit for such long intervals, and I'm not going to allow such severe drawdowns. Why? And I'm going to take my time to transfer the next modules from MathCAD. It's not that I showed results of testing the forecast model on every bar in MathCAD, and they are really impressive.
You may ask, what did I want to show? Just once Saulander promised to test it in MT, I'm fulfilling my promises - here are the first results, which probably must be regarded as "philosophical". :о) Also, I'd like to remind that the Hurst exponent is not such a bad statistic and indeed has a "predictive" property, unlike, for example, wavelet transform coefficients :o).
PS1: And you don't need, not two not three, not one hundred channels - it's enough to have one and you can already get results. Well... whatever you have.
PS2: "...the 15th tortured for 10 months. Why in the 15th trade the channel showed SELL, and not BUY? I guess Chaos is to blame. It certainly hurts that the system patiently waited for 10 months until the price returned where it was supposed to. But the matter can be solved. :о))).
And in this case the conditions of the experiment do not allow us to consider these results as evidence of a statistical advantage.
Hi Sergei !
This test, unfortunately, cannot be considered indicative. In the absence of stops such a ratio of profitable to losing trades is in order. But the absence of stops is a mistake of beginners that is constantly criticized on this forum and completely deprives the results of objectivity.
Try to run the same test in "all ticks" mode with stops. Try to optimize the SL parameter. If you get a positive profit of the strategy with SL<MO, this result will already mean something.
Hello Yuri!
Thanks for the critique. I know that this test cannot be considered as indicative and I know that stops are useful. As I wrote earlier - it's a philosophical test, checking the model, roughly speaking, and most importantly checking it in MT. The vast majority of trades are within reasonable expectation and drawdown. Yuri, only the Hurst index "works".
I attach a test with all ticks (without stops), the results are slightly different. By the way, why is the simulation quality 20%, is there any way to improve it when testing or is there a limit for it?
http://grasn.narod.ru/TEST/StrategyTester1.htm
I won't introduce stops yet - it's in the next implementation, when I add some modules (at least not today :o). As I wrote, the point of testing is not to start trading but to test the model. I might take your advice though. Maybe their introduction will make the implementation of additional modules impractical.
PS: I tested it not only on Eurusd. The results are similar, it works.