Trying to distinguish between real currency mark-up and illusory currency mark-up and recognising the simultaneity of the pairs' moves. - page 7
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
Below I will write in small print, for the topicstarter did not understand this nonsense. If he doesn't mind - let's continue raving.
Here we have again - TI, which few people want to hear about :( But now it is even more twisted: there are already two different entropies - thermodynamic and informational.
To the question of formalization: I don't know myself yet. If we define a macrosignature as a dumb sum of codes (for example, for <+1,0,-1,0,+1,-1,0,0,+1> it is 1), then it is logical to consider as a trend such microstate, which is a representative of a macrosignature macrosignature that is significantly different from zero.
One more time about the terms:
A microstate is a vector in which all currency codes are ordered by all pairs. It is for example <+1,0,-1,0,+1,-1,0,0,+1>. The signature of this microstate is {-2,4,+3}, i.e. two -1, four 0 and three +1.
A microstate always corresponds to the real situation on the market and vice versa. The probability of absolutely any microstate is a constant, due to the very way the codes are calculated (by quantiles), and it equals 3^(-9) for nine pairs. But we all know very well that a strong trend and an ordinary flat are not at all equally probable states. So we have to move to macro states to count the probabilities.
A macro state is the totality of all micro states with identical (=equivalent) signatures. There are a lot of them, but I can cite a few of them to make it clear:
<+1,0,+1,0,+1,-1,0,0,-1>
<0,0,0,0,-1,-1,+1,+1,+1>
<+1,+1,+1,-1,-1,0,0,0,0,0> etc.
The macrostate is entirely determined by the signature, unlike the microstate. The macrostate is needed to estimate probabilities (that's how it seems to me).
Transitions between equivalent microstates (i.e. within the same macrostate, with signature preserved) are free and unpredictable.
Transitions between macrostates are also unpredictable, though not so free.
There is a hypothesis that the thermodynamic entropy of the macrostate is a characteristic tending to inertia, i.e. it may make sense to talk about it in the sense of prediction.
Here is one of the strangest examples of entropy (thermodynamic) inertia: sometimes before very strong news like trailers there is an extremely hard flat (dead calm), when almost all pairs fluctuate within some pips. Suppose at some moment this is a microstate <0,0,0,0,0,+1,0,0> to which corresponds a macrostate with macrosignature {-0,8,+1}. This is a very unlikely macrostate with a probability 130 times smaller than the probability of a typical flat {-3,3,+3}. Its t/d entropy (the logarithm of the number of equivalent microstates) is ln( 9!/(0!*8!*1!) ) = ln(9) ~ 2.20.
After the news is announced, the market explodes on the quid (let's say the quid goes up). This is, let's say, the microstate <+1,+1,+1,+1,+1,+1,0,+1,+1>. It corresponds to a macro-state {-0,1,+8} whose t/d entropy is ln( 9!/(0!*1!*8!) ) = ln(9) ~ 2.20, i.e. the same!
T/d entropy has not changed at all, although the nature of the market has changed dramatically. But something has changed: it is the macrosignature. It was close to zero (dead calm) and now is 8.
I think it would be very interesting to see how the micro-state of the system changes over time. Maybe there are some patterns.
On the subject of formalisation: I don't know yet myself. If we define a macrosignature as a dumb sum of codes (for example, for <+1,0,-1,0,+1,-1,0,0,+1> it is 1), then it is logical to consider as a trend a microstate that is representative of a macrosignature macrostate with a macrosignature that is significantly different from zero.
Alexey, why initially allocate discrete signals +1/-1/0, since it already predetermines/restricts future strategies? Such selection requires a reference point, threshold of triggering - in general it's already some kind of algorithm.
That is, in fact, you are trying to introduce one price action signal that will be checked on each currency pair and detect the current situation of a currency as a vector of these signals, or its convolution. And there are 3 signal values: up signal, down signal and no signal.
This is part of a specific trading system, i.e. it is a specific solution. And signals in general can be different (or with different thresholds) for different pairs. As well as operations with these signals. Of course, the simplest variant is to add them all together and when the sum exceeds some threshold we have a new macrosignal, though there can be other variants.
My post is not an objection to your formalization, but a clarification that we are talking about a specific system. And there may be many variants and only 2 things can confirm the validity of a particular one: statistics and trading logic. If everything is clear with statistics, then it is more complicated with trading logic. What are the processes behind the inertia or reversibility of indices, and respectively, under what conditions can they be expected? After all, it is naive to believe that we can create a perfectly reversible index, whose graph will be a perpetual flat in a narrow range, just like the super-trend version. Ultimately, multicurrency trading comes down to the construction of new synthetic instruments with good properties and the application of trading systems to them.
It looks like this in the picture, but it eats up your computer's resources.
Avals: Алексей, а зачем изначально выделять дискретные сигналы +1/-1/0, ведь это уже предпоределяет/ограничивает будущие стратегии? Такое выделение требует точки отсчёта, порога срабатывания - в общем это уже какой-то алгоритм.
That is, you are essentially trying to introduce one price action signal that will be checked on each currency pair and determine the current situation of the currency as a vector of these signals, or its convolution. And there are 3 signal values: up signal, down signal and no signal.
Yes, the thresholds are defined unambiguously by the p.d.f. of the pair. It is sufficient to divide the distribution into 3 quantiles. The lower and upper thresholds are respectively quantiles q_0.333 and q_0.667. They are not necessarily equal in module and float quite nicely with time.
Yes, this is already some kind of algorithm, and the lack of signal is some kind of improvement on what Semenych suggests. But without quantification I don't know how yet. I was just looking for a clear statistical criterion that group motion has started. By the way, the threshold quantiles are pretty close to zero, on the order of a few 4-digit points on the clocks. In other words, if all the chifs (for example) moved up the chif by 1 point in the last hour, the vast majority of people don't see that group movement as the start of a trend. It's noise. But if they all move 5 pips (in an hour), it is almost certainly a strong group attack on the chiff. That's when you need to enter on the movement. Not against it, as Semenych recommends, but namely against it.
5 pips on each pair seems too weak a move, but it's not. This is where the statistics come in and say that this is already a very unlikely group move, much less likely than a normal flat.
It is already part of a particular trading system i.e. a private decision. And the signals in general may be different (or with different thresholds) for different pairs. As well as operations with these signals. Certainly, the simplest variant is to add them all together and when sum exceeds some threshold we have a new macrosignal, though there can be other variants.
Yes, the simplest one, that is what Semenych suggests - and it is the summation that I don't like. It does not make clear statistical sense. More precisely, thermodynamic. And Semenych is also ironing everything with a wand. I don't like this at all. The iron, of course, makes everything smooth and beautiful, but it also destroys valuable information. It is much harder, but also more fruitful, to find some inert market characteristic that is not an iron.
If the statistics are clear, the trading logic is more difficult.
Absolutely. There are a lot more unpleasant nuances in trading logic.
What are the processes behind the inertia or reversibility of indices and consequently under what conditions should they be expected? After all, it is naive to believe that a perfectly reversible index will produce a perpetual flat, in a narrow range, just like a super-trendy variant. In the end, multicurrency trading comes down to creating new synthetic instruments with good characteristics and applying trading systems to them.
This is something out of a different realm as far as I am concerned. I do not build synthetics. Supertrend synthetics is my other secret project:) I've already given up on synthetic returns: their tails are too thick.
I have to go to bed, I have to get up early. I'll think on it.
2 ULAD: T101, is it?
2 ULAD: T101, is it?
No.
After googling it, I know what it is.
The idea itself came up a year ago. I like it. I'm developing it.
After reading your post I thought you stole my ideas. (Laughing)
I got an interesting one (euromax charts for the same period as above) but it is not ready yet - I will show it to you when I finish it
I would like to check it on ticks. but it's OK, i'll try it on minute charts first
Volumes = a thing of the abstract. It's a pity, but you can't divide them into transaction lines.
Is there any way to do it?
Is there any way to do it?