Market etiquette or good manners in a minefield - page 103
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
1. Here it is actually:
......................
2.Below I have attached a text file with rows of transactions.
The format of the file is as follows:
1. Thank you! It's for EURUSD, M1, PRICE_OPEN, as I understood it // Correct me if I'm wrong.
For comparison, may I draw a couple of other pairs in the same way? Ideally, it would be a closed arbitrage loop
to make a closed arbitrage loop. I.e. EURUSD (already exists)-EURGBP-GBPUSD.
// If it is difficult - don't.
The chart is interesting, I was expecting something else. It's frustrating in itself that
the difference H from 2 is very small everywhere.
2. ... and from the second line down the transactions themselves. It's understandable. Just can't get into the coding.
Working hypothesis:
1) all transactions are continuous, end of one = start of the next
2) transaction sign (buy-sell) is fixed at all, we just watch the result (in points) at the end
3) and is added to the previous result
Or... how?
Sergiy, may I chat with you in Skype? There's a very long and tedious need to tell and explain all my thoughts
to explain, and many of them don't fit in the topic. I would be glad to. And if we come up with something sensible.
we'll come up with, we'll write it up later on the thread.
Shit, I've got the wrong file. I wanted a series of Kagi splits, but I got the RT!
Corrected (see file below).
1. Thanks! It's for EURUSD, M1, PRICE_OPEN, as I understand it.
No, these are ticks for EURUSD.
I do not understand the coding.
*.prn is a text format that can be read with any notepad. If you have a problem, just convert *.prn to *.txt and look it up any way you want.
1) all transactions are continuous, end of one = start of the next
2) transaction sign (buy-sell) is fixed at all, just watch the result (in pips) at the end
3) and is added to the previous result
Or... how?
1. Yes. Closing the current position is opening the opposite direction. RT shows only entry/exit points without taking into account the direction of the opened position.
2,3 I do not understand...
Sergiy, may I have a chat with you on Skype? Here
I have neither the time nor the energy for a private chat. If the topic is secret, then so be it. If not, then it is not me who does not understand something, and it is already a diagnosis.
In other words, the "smoothing" effect for PT is observed only on such H, where the market is arbitrage-free (average Kagi segments tend to 2H), and this may serve as a signal to move to other trading horizons on the parameter H. Or, find a way to exploit this pattern.
The definitions are solved, the entities are almost solved.
It is necessary, imho, to specify at once that the absence of FC (in your opinion)
in the given approach the price is the absence of forecasting between PT extrema
I also suggest that the subject of the prediction should be precisely defined:
What do you propose to predict?
About volatility.
Sergey, pay attention that MetaDriver already focuses on intraday volatility.
The dependence of volatility on the H value attached by you needs to be complemented by the third axis - time, for example in hours of the day.
Look at my charts and the table attached above.
Or/and in your attached files with PT and kagi add for each segment its start time and, if possible, duration (may be of interest).
About the forecast itself.
For the purposes of this topic, I see two forecasting options
(and one more clarifying question: N volatility of what: kagi or PT? in theory it is possible to try the latter)
Let's define exactly WHAT we are forecasting and HOW using the first PT difference you suggest.
Let's define exactly WHAT we predict and HOW we predict using the first PT difference you suggest.
Let's take another look at the Kagi-partitioning of the initial BP:
According to the TS described in Pastukhov's work, positions are opened/closed at each RT countdown (black in the picture). Moreover, the direction of the open position coincides with the direction of the Cagi segment for the n+ strategy and is opposite for the N- strategy. The profitability (average number of points per transaction) for such TS can be estimated as Hvol-2. Let us compare the results of "real" trading with the estimated profitability:
Fig. the red broken line shows evaluation for H+ strategy, the red line with dots shows real trading. You can see that the yield is below the plinth - less than 0.1 point per transaction. This has motivated the search for a more advanced strategy than the trivial one. For example, if we look at the PT in the picture above, we see that it is not strictly sign-variable, as required by the trivial TS. So I had an idea to predict the direction of expected movement of PT using NS by analyzing history for this series. In this formulation, we do not need the initial series of Kagi-divisions (RK) for the analysis, the PT is enough. Moreover, having opened a position we do not need to close it at each RT, which saves the spread. Indeed, if the next PT measurement is expected to move in the direction of an open position, there is no sense to reopen it.
As an example, I have plotted the performance of a strategy that always opens a position in the opposite direction from the previous Pt movement. This algorithm corresponds to an optimal one-way binary pattern. Above I gave estimates of returns for binary patterns built on the first difference of the PT. There is a belief that this analysis is optimal for TS exploiting stationary states of the market in the sense of profitability maximization.
According to the TS described in Pastukhov's work, positions are opened/closed on each count of the PT (black in the picture). The direction of the open position coincides with the direction of the Cagi segment for the N+ strategy and opposite for the N- strategy.
YES
The profitability (average number of points per transaction) for such TS can be estimated as Hvol-2. Let's compare the results of "real" trades with the estimated profitability:
NO, as according to P astuhov it is necessary to be ALWAYS in the trade, which means the profit per one transaction will = H*(N-2) - Spread, where N - N volatility, 2 - "costs" of identification of kagi extremums, i.e. for N=2.1 the profit will be only 0.1*H - Spread
The red broken line in the fig. shows evaluation for H+ strategy and the red line with dots shows real trading. You can see that the profitability is below the base line - less than 0.1 point per transaction. This is what motivated me to look for a more advanced strategy compared to the trivial one.
Imho, you need to recalculate the profit using the formula H*(N-2) - Spread
For example, if we consider the PT in the picture above, we can see that it is not strictly sign-variable, as required by the trivial TS. So I had an idea to predict the direction of expected PT movement using NS by analyzing the history for this series.
i.e. it is proposed to PREVENT the RT on the basis of the RT pattern
In this formulation, we don't need the original Kagi-partitioning (KP) series for analysis, the PT is sufficient.
For example, in the figure I have shown the black line profitability of a strategy that always opens a position in the opposite direction from the previous PT movement.
You yourself write "...if you look at the PT in the figure above, you can see that it is not strictly sign-variable...".
Then on what basis "...opens a position always in the opposite direction from the previous PT movement..."? ? How is it any better than flipping a coin?
This algorithm corresponds to an optimal single-link binary pattern. Above I gave estimates of returns for binary patterns built on the first difference of the PT. There is a belief that for the TS exploiting stationary market states this analysis is optimal in the sense of profitability maximization.
NS, patterns, etc. are just tools. It's known that if the most perfect tool is fed with rubbish at the input, it will be the same at the output.
Now here is a constructive thing :)
Look what, in my opinion, Pastukhov does and he does it, in my opinion, very skillfully.
1. He suggests a new value of volatility.
2. Shows that H volatility is the most STATISTABILY close to 2
3. Concludes that volatility so close to 2 can be used for forecasting
(i.e., is a Reasonable choice of a predictor for stability reasons)
4. Suggests
a. A simple model based on +/- H strategy, KNOWING that volatility is stable
b. A model with patterns, again, KNOWING that volatility is stable
If such order of predictor and forecasting is true and volatility H is really stable with N close to 2, then the following order is true, imho:
1. find N, which gives volatility closest to 2
2. build kagi
3. build PT
And then it is indeed possible to "...open a position always in the opposite direction from the previous PT movement...", since the PT will be as statistically stable as the kagi, and SIGNIFICANT.
Will there even be a difference between the two in terms of valuations?
But smooth, as we know, only on paper,
in practice N volatility "floats" within a day.
Hence
we need to learn how to predict H volatility in order to:
1. for simple model
a. or change H
b. or change the rule of trade opening direction at the beginning of RT
2. for a model with patterns to be sure in predictor reliability
If what is written is correct, attention, a repeat of the question posed a few pages above:
WHAT NON-STATISTICAL METHODS ARE AVAILABLE TO PREDICT VOLATILITY?
The profitability (average number of pips per transaction) for such a TS can be estimated as Hvol-2. Let's compare the results of "real" trades with the profitability estimate:
NONE, because according to P astuhov needs to be in the trade at all times, which means that the yield per trade will = H*(N-2) - Spread, where N - H volatility, 2 - "costs" for the detection of cage extremums, ie for N = 2.1 the yield will be only 0.1*H - Spread
My Hvol-2 is identically equal to yours N-2 , so you are right, I got confused! Right, you need to multiply by N, and without taking into account the commission of DC, the yield will look like this:
I.e. we propose to PREPARE Pt based on Pt pattern
Yes. That seems reasonable.
You write yourself "... if we consider a Pt in the picture above, we see that it is not strictly variable...".
Then on what basis "...opens position always in opposite direction from previous PT movement..."? ? How is that better than flipping a coin?
Yes, for the simple reason that the analysis of binary patterns, the content of which I cited above, speaks (screams) in favor of alternating combinations.
NS, patterns etc are just tools. It's well known that if the most perfect tool is given rubbish at the input, it will be the same at the output.
I agree 100%.
But why all of a sudden you allow the results of statanalysis to be classified as "rubbish"?
See what I think Pastukhov is doing and I think he is doing an abs. competent job.1. it suggests some new value of volatility.
2. Shows that the H volatility is NATURALLY STABILIZED to be close to 2.
3. Concludes that volatility so close to 2 can be used for forecasting
(i.e., is a Reasonable choice of a predictor for stability reasons)
1. Yes
2. Yes
Volatility close to 2 is the essence of a process which is close to a Wiener process, i.e. cannot be used to make statistically reliable profit. What do you want to understand?
Pastukhov, on the contrary, showed the stationarity of the process different from 2, i.e. the arbitrability of the market as such, albeit slightly different from an efficient market.
Do you agree with my interpretation?
You yourself write "...if we look at the PT in the figure above, we can see that it is not strictly sign-variable...".
Then on what basis "...opens position always in opposite direction from previous PT movement..."? ? How is that better than flipping a coin?
Yes, for the simple reason that the analysis of binary patterns whose content I cited above says (screams) in favor of sign alternating combinations
Imho, these are not the results of the analysis, but a consequence of the fact that PT is built on a familiar variable kagi.
I don't want to say it's an accident, but that it's a stability-tested predictor you haven't shown anywhere
Yes, for the simple reason that binary pattern analysis, the contents of which I cited above, speak (shout) in favour of sign-variable combinations.
NS, patterns etc are just tools. It's well known that if the most perfect tool is given rubbish at the input, it will be the same at the output.
I agree 100%.
But why suddenly do you allow the results of statanalysis to be classified as "rubbish"?
I think the easiest way (without shouting :) ) to show not the frequency of occurrence of sign-variable RT patterns, but precisely the stability of their occurrence is to compare MO and RMS.
I propose to do this to clarify the reason for the low yield, i.e. essentially low support for the rule of familiarity of PTs.
See what I think Pastukhov is doing, and I think he is doing it very well.1. It proposes some new value of H volatility
2. Shows that H volatility is NATURALLY STABILIZED to be close to 2.
3. Concludes that volatility so close to 2 can be used for forecasting
(i.e., is a Reasonable choice of a predictor for stability reasons)
1. Yes
2. Yes
Volatility close to 2 is the essence of a process which is close to a Wiener process, i.e. cannot be used to make statistically reliable profit. What do you want to understand?
Pastukhov, on the contrary, showed the stationarity of the process different from 2, i.e. the arbitrability of the market as such, albeit slightly different from an efficient market.
Do you agree with my interpretation?
I meant NOT the EQUATION of 2, but exactly the DILIGENCE of the proximity, i.e. volatility 2.1 is more statistically stable than 2.8.
This is exactly the difference between the Cherkizovsky and Kutuzovsky market models:
- yes at 2.1 we earn less per transaction
- but more often and more steadily
than in the boutique on Kutuzovsky at 2.8.
So, imho, we are back to the same question again:
How to dynamically choose H within a day to get volatility closest to 2 ?
I do not mean EQUALITY 2, but the DILIGENCE of proximity, i.e. volatility 2.1 is more statistically stable than 2.8.
This is exactly the difference between the Cherkizovsky and Kutuzovsky market models:
- yes at 2.1 we earn less per transaction
- but more often and more steadily
than in the boutique on Kutuzovsky at 2.8
So, imho, we are back to the same question again:
How to dynamically choose H within a day to get volatility closest to 2 ?
Stability in this matter is of little value, as it almost reduces everything to "average room temperature".
Especially since H closest to 2 is precisely the least profitable in individual deals. I'm more interested in acquiring
a set of thermometers. :) For me the statistics on H-volatility is like a strategic picture showing
Theoretical profitability (attractiveness) of the instrument and choice of strategies (trend/counter-trend) at different scales.
Metaphorically, it may be compared to the choice of positions for the troops - "tanks to the left, infantry to the right, artillery over there
hollow...". Trading directly on the kagi - God forbid. :) I have more subtle tools for that.
In the image, I've circled the zones that I find interesting. The yellow zone is too retarded, it's for weeklies
and "monthlies", I'm not one of them. :)
The ones, which are too close to 2 are also interesting, but only theoretically (like the zeros of the function).
Intraday dynamics of H-volatility can be defined only for small H (not more than 20-30 at any rate).
There simply are not enough actual data for more. It is possible to calculate an indicator. I am thinking about it. Outputting 8 lines
(MT-limit) for 8 H, respectively, with each line displaying statistical value H-Vol for specified
in the parameters. By the degree of kinkiness of these lines it will be possible to assess the non-/stability of H-Vol.
1) Shit, I've got the wrong file. Wanted a series of Kagi splits, but got the RT!
Corrected it (see file below).
2) No, these are EURUSD ticks.
3)
Just can't figure out the encoding.
*.prn is a text format looks like any notepad. If you have a problem, just rewrite *.prn to *.txt and look it up with whatever you want.
3а)
1) all transactions are continuous, end of one = start of the next
2) transaction sign (buy-sell) is fixed at all, just watch the result (in pips) at the end
3) and is added to the previous result
Or... how?
1. Yes. Closing the current position is opening the opposite direction. PT shows only entry/exit points
entry/exit points without taking into account the direction of the opened position.
2,3 do not understand...
4) I have neither the time nor the energy for private communication. If the topic is secret, then let it stay that way. If not, then it's not me who doesn't understand something, and this is a diagnosis.
1) Whatever, really. One thing translates into another with one bit of extra information :)
2) OK, thanks, got it.
3) :) I didn't express myself correctly, an ambiguity arose. No problem with the physical coding, I got it in three seconds.
I figured it out. But with the logical problems arose about which I asked clarifying questions - (3a).
3a) with 1. correctly solved.
> 2,3 didn't get it right...
well... I had two possible interpretations, the correct one turned out to be the second - which you described. thanks, it's all clear now.
4) OK, got it. Well, if you ever think about it, no problem, come on in. The addresses are above on this thread (page 95). I didn't want to discuss anything particularly
I didn't want to discuss anything particularly secret, just to mutually clear up some questions and to twist some ideas so-and-so.
I've been following this topic for a long time, trying to figure it out, but somehow it's not that simple.
Please enlighten me on what you call RTs and where do you get them from?
Is this what Pastukhov draws in white dots on the kagi chart?