a trading strategy based on Elliott Wave Theory - page 220

New comment
 
Hello everyone!
No sleeping!
I suggest we start discussing Pastukhov's thesis.
http://forex.kbpauk.ru/showflat.php/Cat/0/Number/142956/an/0/page/0#Post142956
There is one question on the agenda: about the advantage of the transition from the traditional horizontal breakdown of time series into TFs to the vertical breakdown into discrete candles of the same amplitude.
I show you the results of my research for introductory purposes. To estimate the method profitability standard tools were used: FAC and standard deviation. Input data was tick quotes for the whole 2006 for EURUSD, EURCHF and EURGBP. The initial time series were sliced according to the method proposed in the thesis, and a new synthetic series with bars having the same amplitude(A) in points (respectively, with different lifetime of each candle) was generated. Synthetic series were created in increments of one point, from 1 to 50 points and for each one the estimation of returns was made:
s(A)=|FAC|*A, where A is the discretisation step on the vertical axis.
and an estimate of the total return for the year:
S(A)=const*(s(A)-Spread)/A^2
The average yield per trade is presented in the picture. The averaging was performed on a period of one year, the number of events in the worst case is not less than 50.

What attracts attention is that the scheme allows exceeding profitability of spreads that exist nowadays. That is good!
In the picture below we can compare the integral profits of the scheme for different instruments.

As we can see, the instrument returns are generally comparable to each other and lie in the range of several hundreds of points per year at a spread of 1 point for EURUSD and 2 points for EURCHF, EURGBP. For EURUSD the earning omimum lies in the area of 35-point candlesticks.
Generally speaking, the idea shown in the dissertation is not new; the point is that by using TP and SL orders in our TS we, in fact, implement the proposed trading scheme. Indeed, in its simplest form, our strategy will be reduced to placing equal in value TP and SL orders and opening a position in the direction of the previous movement, if the FAC is positive, and in the opposite direction, if it is negative (the author uses the Hurst-N indicator, which does not change the essence). The procedure of opening a new position is executed immediately after the current one is closed. By the way, using unequal ТР and SL, you can significantly increase the strategy efficiency. And TP> SL or vice versa will directly depend on Hurst ratio.
In such an ingenuous way it is possible to realize what is well written about in a 100-page dissertation and for which a scientific degree of Candidate of Science in Physics and Mathematics has been awarded :-)
If we compare the proposed method of conducting arbitration deals with classical methods, we can note
1. comparable integral returns. Which is neutral.
2. Smaller capitalization of the opened position with the same market risk due to larger TP and SL orders. Which is not good!
3. Only ONE parameter of the TS optimization - the amplitude of candles. This is very, very good. You can guarantee the robustness of the method.
 
double post
 
Neutron 16.01.07 11:19
I have a sneaking suspicion that you've been reading the wrong dissertation. :)

Actually, there are two methods of "partitioning" shown there, kagi and renko. Even though they're considered
in parallel, they are still different.
The author does not use the Hearst index, he only refers to it saying that the H-volatility introduced by him
is structurally similar to Hearst's index.
On the basis of the H-volatility the performance of some trading strategies (described in this dissertation) has been suggested and proved.
trading strategies (described in the dissertation).
It is also proved that H-division optimally (in some sense) describes the initial data series.
(conventionally speaking, it can be used as a "noise suppressor"). This is the only thing that you
used, of the whole thesis.

PS/ but that's just me, not picking on minutiae, just wanted accuracy.
 
Stay awake! <br/ translate="no"> I suggest we start discussing Pastukhov's thesis.


I support it in general. On one condition, however - not to forget about my own "dissertations" too.
. Therefore I publish the picture, which I should have done long ago.
Since I haven't installed the matcad yet, the picture is in eexcel. don't scold too much.
This symbol is GBPUSD, M15
Amount of bars on the chart - 38856
Potential entry points - 4038
Blue points - successful entry, red - unsuccessful
Criterion - entry was considered successful if it resulted in +6 points or more (correction for spread),
others, including positive ones, were unsuccessful.
Total number of successful entries - 3482, unsuccessful - 556
From my point of view, we can say the only thing about this picture, that using the phase plane
of the indicators BR,BL does not allow to build a criterion for dividing the set
of deals into successful and unsuccessful ones.


PS I support Northwind's initiative to publish pictures and text in
screen format. I hope I won't have to fidget back and forth with the screen when reading this post. :-)
 
Yurixx 16.01.07 13:40

...Total of successful inputs - 3482, unsuccessful inputs - 556
From my point of view, the only thing that can be said about this picture is that the use of phase
of indicators BR,BL does not allow to build a criterion for dividing the set of
into successful and unsuccessful deals.

What do you mean you cannot build a criterion? You have the ratio of successes to failures as 7 to 1, already.
And what is shown in the picture allows you to construct something like BR=a+b*BL.
But that's not what you want.
 
By the way, I haven't read everything yet, but it seems that the author doesn't really use the Hearst value. He introduces the H value, but redefines this value. Maybe, of course, I haven't finished reading it yet...:o)
 
Pastukhov's dissertation at the link above is a broken archive, not opening.
 
What do you mean by failing to build a criterion? You have a success-to-failure ratio like 7 to 1, already.


That's hindsight. It's not like I'm running a strategy on a tester. I just have the same indicator whose signals
I need to filter out. On history these signals are known, in real time to identify the signal
takes time. If this lag time is taken into account, perhaps this advantage will disappear completely.
At best, it will be significantly reduced.
 
Pastukhov's dissertation at the above link is a broken archive, doesn't open. <br / translate="no">

Strange, everyone opens it, you don't. Why is it broken?
1...213214215216217218219220221222223224225226227...309
New comment