Discussion of article "A scientific approach to the development of trading algorithms" - page 6

New comment
 
Rustem Bigeev:

This is how a regularity was found: markets tend to be trending on any scale, but as the scale increases, the trending decreases, i.e. after passing N points vertically, with a probability of more than 50%, the price will pass the same number of points in the same direction. This pattern is good because it allows you to use a simple trend strategy for trading, where you can open a Buy position after each upward move and open a Sell position after each downward move.


Do you not take into account the situation when the price starts to run through several of your blocks in one tick? In this case your blocks are drawn post facto and you cannot open trades on them.
Of course I do. That is why the scale is taken large, for example, the candle size*10...14. The article describes this moment, I called it the delta between the closing price of the block and the candle. There may be situations when several blocks will be formed at once, but they will be not so frequent. But the point is that knowing the problem, it should be solved, consistently smaller delta.
 

Actually, L.A. Shelepin showed that the first difference alone is not enough for a stable forecast of the next step (see the attached article).

And to say that if the price has passed a certain number of points in one direction, the probability of the same next step is >50% - unfortunately, it is not necessary. At any price quantisation, it is necessary to take into account at least the second difference in calculations.

 
Alexander_K:

Actually, even Shelepin L.A. showed that one first difference is absolutely insufficient for stable prediction of the next step (see the attached article).

And to say that if the price has passed a certain number of points in one direction, the probability of the same next step is >50% - unfortunately, it is not necessary. In any price quantisation, it is necessary to take into account at least the second difference in calculations.

In some cases it works, when the asset has high investment attractiveness, the money coming to the market is enough to create a deviation from 50%. Such an algorithm as an alternative to the simple purchase of investment attractive securities has the right to life, because it is able to earn not only on the growth, but also on the fall. And further, even such a simple idea can be developed very much, changing it almost to the point of unrecognisability. It's basically a skeleton.
But of course it is obvious to me that the patterns are much more complicated and it is necessary to take into account not only the last movement, but ideally all movements.
 
Maxim Romanov:
In some cases it works, when an asset has a high investment attractiveness, the money coming to the market is enough to create a deviation from 50%. Such an algorithm as an alternative to the simple purchase of investment attractive securities has the right to life, because it is able to earn not only on growth, but also on the fall. And further, even such a simple idea can be developed very much, changing it almost to the point of unrecognisability. It's basically a skeleton.
But of course it is obvious to me that the patterns are much more complicated and it is necessary to take into account not only the last movement, but ideally all movements.

Well, my comment does not mean that one should instantly abandon research in this direction. Especially since I myself am a supporter of the opinion that it is in the form of distribution of increments, in its "tails", that the Grail sits.

However, if we are talking about "memory" as a source of cash (as opposed to SB), then from the series of your articles this memory of the market is not obvious at all, to the point that only the last increment is investigated, and at least the previous one is not taken into account.

In fact, working only with the last retournment is just implying that you agree that the market has no memory and is a regular Markov process.

I think that if you apply your metaOde to SB, you get exactly the same results. And that's not a good thing.

 

I didn't like the article.

All the reasoning is based on some "regularity" that simply cannot exist.

 
Maxim Romanov:
In some cases it works, when an asset has a high investment attractiveness, the money coming to the market is enough to create a deviation from 50%. Such an algorithm as an alternative to simple purchase of investment attractive securities has the right to life, because it is able to earn not only on growth, but also on the fall. And further, even such a simple idea can be developed very much, changing it almost to the point of unrecognisability. It's basically a skeleton.
But of course it is obvious to me that the patterns are much more complicated and it is necessary to take into account not only the last movement, but ideally all movements.

and now they are also trying to decompose the market into trend phases and flat phases.
to trade only trend phases.


 

A more specific concept of trend is given in the theory of impulse equilibrium.

For this purpose, fundamentally new concepts are introduced there: stimulating amplitude, active impulse, superactive impulse.

Applied to the formation of a trend, these concepts are used as follows:


1. the basis of the trend is the inducing amplitude of the price (the amplitude that affects the interests of the vast majority of market participants).

2. The active impulse is trend-forming (because it is connected with the inducing amplitude).

3. Over-active impulse, on the contrary, is the cause of growth of multidirectional amplitude of movement, volatility,

and, as a consequence, a critical correction or destruction of the existing trend.

4. And the main thing: a trend is always the same section of the elementary structure, regardless of the scale of price movement (called M-shape).

 
Aleksandr Masterskikh:

A more specific concept of trend is given in the impulse equilibrium theory.

For this purpose, fundamentally new concepts are introduced there: inducing amplitude, active impulse, superactive impulse.

When applied to trend formation, these concepts are used as follows:


1. the basis of the trend is the inducing amplitude of the price (the amplitude that affects the interests of the vast majority of market participants).

2. an active impulse is trend-forming (because it is related to the inducing amplitude).

3. Over-active impulse, on the contrary, is the cause of growth of multidirectional amplitude of movement, volatility,

and, as a consequence, a critical correction or destruction of the existing trend.

4. and the main thing: a trend is always the same section of the elementary structure, regardless of the scale of price movement (called M-shape).

Then write here the definition of trend from this theory, if it is really more specific, people will be more interested to familiarise themselves with your theory.
 
Alexander_K:

Well, my comment does not mean that one should instantly abandon research in this direction. Moreover, I myself am a supporter of the opinion that the Grail sits exactly in the form of distribution of increments, in its "tails".

However, if we are talking about "memory" as a source of cash (as opposed to SB), then from the series of your articles this memory of the market is not obvious at all, up to the point that only the last increment is investigated and at least the previous one is not taken into account.

In fact, working only with the last retournment implies that you agree that the market has no memory and is an ordinary Markov process.

I think that if you apply your metaOde to SB, you get exactly the same results. And that's not good.

If you apply it to the SB, the output of the balance graph is the SB, I tried it, because the first thing I had to do was to check if I was wrong.
As for what increment to work with, I went this way: if we take different scales of blocks and take only the last increment, it is an analogue of the study with the previous increment. For example, we take blocks of 5 points, 6, 7, 8, 9, etc. and analyse the last increment. So far.
But this is not my main work, just a piece of it. In my main work I use a sample of 24-40 blocks and analyse its structure, I do such analysis on 5-10 scales simultaneously.
 
multiplicator:

and now they're trying to split the market into trend phases and flat phases.
to trade only the trend phases.


I used to talk to this guy. back when their MC forum existed. I will say they are mathematicians, but their approach to trading is so bad ... only hft were really profitable at them, the rest at the level of mql signals. But they did not offer hft to clients because there is no point.
123456789
New comment