Using Neural Networks in Trading. - page 13
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
In short, it's difficult, if not impossible, to think of anything better than Neuronka for price analysis, unless it's insider information.
This is where it all started once upon a time: studying NS as applied to prices, but it turned out to be hard to get an output in sliding time window mode. So I have to put Haykin, Bishop and Shumsky on the shelf and sit on the Renko scale instead of the time window. I have both direct and indirect evidence for the prospect of this direction. One of that evidence is made by you Neutron, when you subtracted 2H from Kaga's shoulders, pitched to NS and got a pretty good result to start with.
Why Renko and not Kagi?
Kagi splitting is known to be more responsive to changing trends. This is despite the fact that the practical implementation of both is indistinguishable in complexity and takes three lines of code.
Why Renko and not Kagi?
Kagi splitting is known to be more responsive to changing trends. This is despite the fact that the practical implementation in terms of complexity of both are indistinguishable and take three lines of code.
I also thought so, but an interesting nuance appeared (I read it in "Trading with movement" section of Spider): Renko scale, on which the strategy is optimized, must be maintained in further testing, i.e. besides the box size (which is equal to the difference between the scale levels) the scale levels must remain in one place. And what to do with Kagi in this case personally I have not yet decided, but maybe you will succeed.
...the Renko scale must be maintained, i.e. in addition to the size of the box, the scale levels must remain in the same place.
I can't understand it! What are scale levels?
Here, the trap is this - Renko splitting is not absolute, i.e. the result of its effect on the time series (RT) depends on which RT datum to start splitting from. This uncertainty has the same order of magnitude on the price scale as the size of the box. In other words, it is possible to partition BP in several different ways and the maximum difference for Renko is no larger than the partitioning step, H. This disadvantage is absent for Kagi partitioning. Here it's unambiguous, no matter how you start this partitioning.
In short, whichever way you go about it - Kagi-partitioning is cooler!
I couldn't follow it! What are the scale levels?
The catch here is that Renco splitting is not absolute, i.e. its effect on the time series (TP) depends on which TP datum to start splitting at. This uncertainty has the same order of magnitude on the price scale as the size of the box. In other words, it is possible to partition BP in several different ways and the maximum difference for Renko is no larger than the partitioning step, H. This disadvantage is absent for Kagi partitioning. Here it's unambiguous no matter how you start this partitioning.
In short, whichever way you go about it - Kagi-partitioning is cooler!
That's what I wrote, but I guess I didn't quite get it right. Stepped away from the computer and thought about Kagi and decided that it, too, is devoid of this drawback. On the other hand, if Renko takes into account each box and not only its extremums, we know the +/- levels and can place a pending order and not an order with immediate execution. Although, it is possible to try all these situations.
I'm worried about organization of Renko/Kagi constructions in MT4. After all, if we write an Expert Advisor based on these structures, then during initialization it should download the minute history and fill the buffer of the structure and this is the only buffer we work with. But we cannot write a separate Renko/Kagi and access its values, because MT4 has a timeline everywhere. Am I thinking correctly?
http://www.forextrade.ru/?p=3251&id=143267
The other thing is interesting, the H value. In one of the reports (file attached) Shiryaev keeps talking about figure 2 and revolves around it. Who has any opinion ? What is this 2, what is it measured in ?
I just want to check if I am right in calculating this 2
Hi Sergey.
Two, is a relative dimensionless quantity which is the ratio of the average distance (in pips) between the extremes of the partition to the step of partition H (in pips). For a Wiener process this value tends to 2 if n is large. On real market BP this value is always different from 2 and characterizes the degree of non-arbitrage/arbitrability of an instrument.
Hi Sergey.
Two, is a relative dimensionless quantity which is the ratio of the average distance (in pips) between the extremes of the partition to the step of partition H (in pips). For a Wiener process this value tends to 2 if n is large. On a real market BP this value is always different from 2 and describes the degree of nonarbitrage/arbitrage of an instrument.
Strange, this is a random value ((
My calculation is different. It is calculated differently. But I will only confuse you.