Optimal values of SL and TP orders for an arbitrary TS. - 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
Анатолий, дайте пжл определение Вашего понимания "динамической ТС"
Like this (imho): Following the market, ie capable of independently finding the best conditions for action (opening / modification / closing) at the current price series. (here on the forum seems to practice only one option to create such a TS, in which self-optimization is conducted, and automatically by some criteria best options are selected (if I'm wrong, point fingers where necessary (: ).
That's how TC is needed, the method you suggested.
Although it probably can be used for Expert Advisors in which there are several static systems (action parameters are invariable), i.e. for each TS the coefficient is different.
//--
Forgive me for being too curious! Probability Michael, apologize at once excessive curiosity, the calculation technique, required to determine the lot, no doubt a thin (, and interesting, and therefore) could you demonstrate the work of the impact of probability on the lot and the reflection of this interaction on the balance, for example putting two balance charts from the tester with enabled and disabled the probability function, with the same parameters, preferably that history contains both bullish and bearish trends, I would be very grateful
p.s. Suppose to open a trade, the probability must be >=70%; and with a probability of 70% opens a trade with a risk of X% of the deposit, then what will be your risk X, when you open a trade with a maximum probability of 100%?
Somehow (imho): following the market, i.e. having the ability to find the best conditions for action (opening/modification/closing) on the current price series. (seems to be the only variant of such a TS practiced here on the forum, in which auto-optimization is performed, and the best options are selected automatically according to certain criteria (if I am wrong, put your finger here (: )
Thanks for the definition, I hope it will be useful to the topicmaster.
If you are seriously interested in the MM+Markovec topic, PLEASE start a separate thread and state the problem and your options there.
I think it will be interesting for me and many others to participate in the discussion too.
Если Вас всерьез интересует тема "ММ+Марковец", ПОЖАЛУЙСТА, заведите отдельную тему и изложите в ней задачу и свои варианты ее решения.
Думаю, и мне и многим другим будет интересно поучаствовать в обсуждении.
I think the topic should be started by someone who has been digging in this direction and has some results. I just got interested and made a note of it for the future.
grasn писал(а) >>
I had a very simple idea. My reasoning was that if you are looking for a "universal" solution for SL, not related to a specific strategy, then you need to determine the "model" of the market (in quotes). Only in this case, we can hope to find something acceptable (Sl and TP are related - it's a medical fact).
Sergey, I am of the opinion that the task of protective orders should not be reduced to making trading decisions instead of TS.
TS should be built based on the accepted "market model". The purpose of SL & TP orders is either to prevent games of chance or (this is a more subtle point!) to cut off infinite tails in FS for a particular TS which allows reducing the risks of TS to some extent and as consequence increasing the share of deposit f, taking part in trades. It is the latter that determines a clear affiliation of protective orders to a particular TS, without which their consideration in the context of an accepted "market model" is meaningless.
Market model
The market model I have adopted is very simple. It's a "superposition" of two processes: Bernoulli's process ("momentum positive" and "momentum negative") and a very complex distribution of incremental modules, similar to lognormal (from a distance :o)). It works simply - a positive or negative pulse is generated, then it is multiplied (accelerated) by some positive value (including zero) derived from a distribution close to the distribution of market returns. But these retournals are so complicated and their distribution so incomprehensible that I have decided to replace them with some average of these retournals.
PS: Now it's your turn to criticize :o).
Regarding this I have only one remark. If we consider the quoting process of an instrument as a parity between two groups of speculators, it seems to me necessary to introduce into the model the "horizon of interest", which is responsible for the characteristic duration of investing for different groups of investors. It is this phenomenon that determines the tricky law of RRR distribution in the price series that you note. You lack this point and are forced to resort to artificial points in describing features (highlighted in your post). In my opinion, the main determinant of the quoting process for financial instruments is the stabilizing action of the state banks. Exactly for this reason we do not observe the collective effects (trends, panics, we do not observe the positive autocorrelation coefficient in EPR, we do not observe H>2 for Kagi-division, we do not observe H>1/2 for Harst index, etc.) - Forex as a rule has a rollback or in other words, a flat character.
I think that it is impossible to build a market model without knowing its specific laws, which are reliably hidden from us. But we can build a TS that fully takes into account any (or almost any) regularities in the price series. In general, it seems strange - there is no model, but the TS nevertheless exists... We need to discuss all this, that's what this thread is designed to do. I'm asking everyone to be more active in this thread with their ideas on these issues, even if they don't coincide with the topic of the thread.
To clarify, an arbitrary but profitable system. Any SL and TR theory is meaningless without reference to a particular particular particular TS. Another TS will give different values of SL and TR.
Agreed. I said "arbitrary" to emphasize the generality of the reasoning. The criterion of profitability of TS will be obtained within the framework of the accepted model.
Yurixx wrote >>
Hi Serguei!
>> Very interesting. Not to say that it is relevant. And even now, at the early stage of exposition, it is seen how it can be applied in practice. However, maybe you have some other surprises in store for me ? Well, I'm waiting for the sequel.
Hi Yura!
It's very nice of you to show a keen interest in the research you have done. Your comments are always valuable.
PapaYozh wrote (a) >>
Still the starting value K[0]?
You write quite well, it's strange that you didn't write it as an article. Though it's your business, I'll be glad to read the rest.
Thank you, PapaYozh, for the remark. Really, you must read K[0] instead of the stated K[n]. I made a mistake.
Regarding an article on a website, I believe that an article must be a prepared material that has been subjected to obligatory expert censorship and must be ready for use (not raw). For this purpose usually carry out various conferences where the material exposed to public discussion passes necessary check and criticism and only then is issued in the form of articles. Unfortunately, on our site this practice is completely absent, and articles are often posted that contain information which has nothing to do with reality and resembles (not always scientific) delirium of a man sick in the head. It is eerie when there is no professional censorship. Moreover, it is actual for us - traders, because it is no secret, that the field, where the Big money turn over, attracts flies like honey, people, let's say, sometimes not quite adequate and, moreover, the responsibility of this or that statement, if the real money is on the line, is great.
For this reason I prefer to work on a raw material at first on a forum and only then, in case of high quality of presented information, to undertake responsibility to make it out as an article.
This is how it should be done (
imho): It follows the market, i.e. capable of independent search of the best conditions for actions (opening/modification/closing) in the current price series. (It seems to be practiced only one variant of such TS on this forum, in which auto-optimization is performed, and the best options are selected automatically according to certain criteria (if I'm wrong, put your finger here (: )
.)
We don't need to get ahead of ourselves with auto-optimisation of parameters. We will obtain in analytical form an expression that will include fraction of equity f, protective orders SL & TP, spread Sp and parameter responsible for "predictability" of price series in the most general case. All these parameters are functions of predictability and the specific type of the adopted strategy (which is determined by the type of TS bribes - its FR). Having such an apparatus in hand, it is not difficult to investigate the prospect of introducing dynamic parameters. Therefore I propose to first build a basic consistent design where everything falls into place and only then start playing with parameters and details. This is the way to a successful understanding of the market. Otherwise we will bury ourselves in a heap of unimportant or unimportant details without seeing the main point. This is exactly the case which is implemented in MTS optimizer - we come up with something and sit on top of the endless problem of exponential growth of task's complexity with increasing number of optimized parameters (not necessarily relevant). This is the reason why I'm trying to move towards an analytical solution to the problem. In this case, we potentially benefit from the fact that the justification of a logically accepted structure to us comes to understand the processes that determine some of the moments in the trade and open up new sides of the building called the Market.
Сергей, я придерживаюсь мнения, что задача защитных ордеров не должна сводится к принятию торговых решений вместо ТС.
ТС должна строится исходя из принятой "модели рынка".
But I did not say anything about protective orders. This is a separate topic and I haven't considered it yet due to its serious complexity.
The task of SL & TP orders, on the other hand, is either to prevent force majeure, or (this is a more subtle point!) to cut off endless tails in the FS for a particular TS, which allows to somewhat reduce TS risks and, consequently, increase the share of the deposit f involved in trading.
Then a question for you, what does TC do, if it does not define TP or SL? My conception is very simple. TS determines the TP level, which should be expected within a certain period of time. If the price moved in the other direction and reached a certain level, then most likely after this event (with a high degree of certainty) will not go back - to the level of TP. This critical level I call SL (non-return level at some time). Approaching to this level and moreover passing through it is a forecast error. I've also come to the same unambiguous conviction - if you don't feel the "time" in the market - there is nothing to do there. Moreover the market is the time.) That's the philosophy.
I suspect that your TS assumes that (for example) it starts to grow, opens an order and every bar/trick, checks if the growth continues? Until some criterion confirms otherwise. In this case, apparently, the system does not understand what TP to expect and whether this growth will continue in the next iteration. This is typical of many systems, but the approach to TP and SL is only true for this methodology and it is not the final truth. Sergey, there are just other systems and models :o)
Of course, such systems also have the right to life, and perhaps they are good, but I gave up on them for different and very good reasons.
It is the latter that determines the clear affiliation of protective orders to a particular TS, without which their consideration in the context of an accepted "market model" is meaningless.
And what makes you think that TP of any particular trading system will have thick tails? This is an observation or conclusion?
And there is another subtlety, or maybe it's not clear, or maybe it's nothing. The essence of its nonsense is that you "kill" the information in some sense. Let me try to explain. You have separated TP and SL from TC. They live their lives in terms of data (information). In other words, if you're going to take statistics (distribution) by points of profit/loss, then you need to analyze orders closed only by TP/SL and closed by TS at all. And we should use in the formula only the distribution which is closed by TP/SL. But how will it all relate to the TS?
Regarding the above I have only one remark. If we consider quoting of an instrument as a parity between two groups of speculators, it seems to me necessary to introduce into the model "horizon of interest" which is responsible for characteristic duration of investing of funds for different groups of investors. It is this phenomenon that determines the tricky law of RRR distribution in the price series that you note.
I don't really understand what speculative groups have to do with it. If you mean "bulls" and "bears" and more subtle classifications, then I'm sorry - I do not understand it. All the indicators and "methods" that can separate one from the other, like flies from cutlets, I consider a complete nonsense - I'm being honest and straightforward.
You lack this point and have to resort to artificial points in describing the features (highlighted in your post).
The "method of maximum likelihood", which, of course, is artificial in some sense, but completely workable, and the word "concept" are highlighted there. I don't really understand your comment, so I thought I'd explain it in the abstract:
If you want to know my opinion concerning this question, I consider that the main point determining the quoting process of financial instruments is the stabilization actions of state banks. Exactly for this reason we do not observe the collective effects (trends, panics, we do not observe the positive autocorrelation coefficient in RRR, we do not observe H>2 for Kagi-division, we do not observe H>1/2 for Harst index, etc.) - Forex is usually of rollback or in other words, of flat nature.
Banks are certainly important - you're right there. But it makes no sense to consider and rely on ACF - it says nothing on these rows. It simply says nothing. You cannot get ACF for these rows. You won't get that limit for such nasty rows. But Hers is noticeably higher than 0.5, even taking into account biased estimates, so you're wrong. It just shows the tendency to hold/go away from the average. There is nothing wrong with the quotes - they do not linger on their average (which, due to the thick tails, has to be calculated, as the standard formula does not show any average). But if you are sure that Hearst is 0.5 for quotes, then show us what is the mean value of quotes.
I think that it is impossible to build a market model without knowing its specific laws, which are reliably hidden from us. But we can build a TS that fully takes into account any (or almost any) regularities in the price series. In general, it seems strange - there is no model, but the TS nevertheless exists... We need to discuss all this, that's what this thread is designed to do. I ask everyone to be more active in this thread to put their ideas on these issues, even if they do not coincide with the subject branch.
Let's not do this "squeeze" thing. Any trading system makes assumptions about the market in the form of models. And these models can be very simple or very different - in short, different. Of course not a single model considers all the inexplicable complexity of the market, but the colleagues don't need that, they take only what is potentially profitable, nobody suffers from academia here :o)
I have only one comment on the above. If we see the process of quoting an instrument as a parity between two groups
it is very easy to imagine that "sellers and buyers" are meant, and they are, but such a "simple" gradation will not help us understand the process. i would like to specify that "sellers and buyers" of groups of DIFFERENT INVESTMENT HORIZONTS. i believe this clarification will describe the situation more clearly. and then the PARITT becomes more understandable. (I think this could affect the presentation of the formula.)
If you are a speculator, it seems to me necessary to introduce into the model the "horizon of interest", which is responsible for the characteristic durations of investing for different groups of investors. It is this phenomenon that determines the tricky law of RRR distribution in the price series that you have noticed. You lack this point and are forced to resort to artificial points in describing features (highlighted in your post). In my opinion, the main determinant of the quoting process for financial instruments is the stabilizing action of the state banks. Exactly for this reason we do not observe the collective effects (trends, panics, we do not observe the positive autocorrelation coefficient in EPR, we do not observe H>2 for Kagi-division, we do not observe H>1/2 for Harst index, etc.) - Forex as a rule has a rollback or in other words, a flat character.
I think that it is impossible to build a market model without knowing its specific laws, which are reliably hidden from us. But we can build a TS that fully takes into account any (or almost any) regularities in the price series. In general, it seems strange - there is no model, but the TS nevertheless exists... We need to discuss all this, that's what this thread is designed to do. I ask everyone to be more active in this thread to post their ideas on these issues, even if they do not coincide with the topic of the branch.
We don't need to get ahead of ourselves with autotuning of parameters. We will obtain in analytical form an expression that will include equity f, protective orders SL & TP, spread Sp and the parameter responsible for the "predictability" of the price series in the most general case. All these parameters are functions of predictability and specific type of strategy adopted
I think ONE strategy should be considered only within the framework of OTHER strategy(s), a sort of matryeshka. The consideration of ONE strategy in isolation from the consideration of other strategies makes it unprotected and vulnerable. Like work schedules, strategies must have a "power of action" gradation (like an adult allowing a child to misbehave (or be entrusted with a task) but within the limits of what is allowed). ( stone, scissors, paper?)
... and there will be several formulas... or rather, there should be one formula, but it will live in "different boxes"... each strategy is good at a certain moment, so it is necessary to introduce a "lifetime" of the strategy. it can also be a parameter that excludes (filtering) taking into account "N-strategy"...
(which is determined by the type of takings of the TS - its FR). With such an apparatus in hand, it would not be difficult to investigate the prospect of introducing dynamic parameters. That's why I propose to first build a basic consistent design where everything falls into place and only then start playing with parameters and details. This is the way to a successful understanding of the market. Otherwise we will bury ourselves in a heap of unimportant or unimportant details without seeing the main point. This is exactly the case which is implemented in MTS optimizer - we come up with something and sit on top of the endless problem of exponential growth of task's complexity with increasing number of optimized parameters (not necessarily relevant). This is the reason why I'm trying to move towards an analytical solution to the problem. At the same time, we potentially benefit from the fact that substantiating the logically accepted construction we come to understand the processes that determine some moments in trading and open up new sides of the building called the Market.
Но я ничего не говорил о защитных ордерах. Это вообще отдельная тема и ее пока не рассматривал в виду серьезной сложности.
Тогда вопрос к тебе, чем занимается ТС, если не определяет TP или SL?
I believe that an "ideal" TS should only identify entry/exit points. That's it.
If for some reason the TS was unable to implement an exit (disconnection, etc.), then a protective order is forced to perform this function. This is its main purpose - insurance. However, it is not all that simple. There are TS with long tails in the Take Profit Department, and as a consequence, the probability of losing the deposit during normal operation of such TS cannot be ruled out... This is where the necessity of performing "sacred trimming" of the foreskin of the FR appears :-) The introduction of this parameter worsens the profitability of the TS, but also reduces its extreme risks, i.e. there are two competing processes, which means we need optimization of SL & TP parameters.
Once again. I clearly separate the tasks to be performed by TS and protective orders and this approach allows us to find the optimal solution for TS and SL & TP separately. You seem to confuse these concepts by solving an optimization problem with a smaller number of parameters. Maybe you're right, but I see the means of solving the problem in the formulation I advocate - TS - separately, SL & TP orders - separately.
And what makes you think that the TF of some particular TS will have thick tails? This is an observation or conclusion?
I try to use the most common approach, that's why I do not exclude such an extreme variant.
You have separated TP and SL from the TS. They live their lives in terms of data (information). In other words, if you're going to take statistics (distribution) by points of profit/loss, then you need to analyze the orders closed only by TP/SL and not consider closed TPs at all. And we should use in the formula only the distribution which is closed by TP/SL. But how does this all relate to the TS?
It is easy. Consider for example the same TS with an arbitrary distribution of payoffs. Let this distribution for a TS without protective orders be of gaussian form and have "infinite" tails (shown in red in the figure):
It is clear that such TS with positive MO=10 points will surely lose the deposit sooner or later, because the inevitable thick and extremely negative tail (black swan, Moose) will come. To avoid this, we introduce protective orders in the system that forcefully close a position if its profit or loss value exceeds a certain level. For such a smart trading system, the FR has significantly changed (in blue): thick tails have disappeared and horns on the FR edges have grown. This is inevitable as the frequency of events with protective order parameters has locally increased.
Let me repeat. The profitability will generally decrease for such an TS, but the risks will also decrease which will allow us to increase f. To estimate the influence of orders on the TS operation, we need to modify the integral expression for an ordinary TS obtained on the previous page. I will remind you of it:
For this purpose let's divide the integration segment into three (using the additivity property), each being responsible for its function. The first section will be responsible for the SL-order and will look as follows:
It simply sums up the number of "hits" for SL and thus finds the height of the left horn in the FR of the TS takedowns with the protective orders. This is correct for the right ear:
Well, for the middle one it's simple:
Finally, writing down the expression for the logarithm of profit of the TS with protective orders, we have
Here we have such an expression responsible for money in our pockets when working on Forex. We are going to look for its extremum...
Now it's time to enter the Sp parameter responsible for the DC commission or spread into the analysis.
Later...
Sergey, however, this breakdown into three areas does not take into account one important effect. A part of profitable orders has time to be in the SL zone before they close (the same is with TP). It means that introduction of SL (as well as TP) deforms considerably the entire distribution, including the zone of standard TP exits. Or I am getting ahead of myself and further you will describe this effect?
In general, I must say that I read the material with great pleasure, but you asked for criticism :)
to Neutron
Я считаю, что "идеальная" ТС должна определять только точки входа/выхода. Всё.
But TP and SL are part of the "entry/exit" system - they are exit options, one of them by profit, and since the parameter "profit" appears, in a sense it is possible to think about profit optimisation as a minimum. If there is no profit in explicit form, then it is probably possible to optimize by it, but it is a bit complicated. (I'm not talking about the MT Optimizer itself).
Once again. I have a clear division of tasks that TS and protective orders should perform and this approach allows us to find the optimal solution for TS and SL & TP separately.
Serega, I have no doubt you clearly distinguish them. I would have to do the same for others, since you opened the topic. I just do not know all about your approaches and I can not assess for sure, whether your method is suitable or will have to invent their own for their purposes. So I'm asking you, for the simple reason that you have only now started to make them clear to me. I meet some mentions of correspondence, Seryoga, I am not a telepath, and not a mathematician and you do not get angry. Although I admit that I alone do not understand the "difference" indicated somewhere earlier :o(.
I don't understand the "difference" between the order and the TP and the protective order.
Let's go over it again, but specifically define the terms. Your TS opens and closes orders by itself. But exactly for each order you open a protective order? This is a huge difference, it is additional - the opening price, SL, TP, spread, lot, direction (and not necessarily the opposite). And what you have written should be attributed to protective orders?
Is it true or it is just a term and we are speaking about the level of TP, SL and lot for one current order?
That's easy. Let us consider one and the same TS with an arbitrary distribution of bribes as an example. Let this distribution for an TS without protective orders be of gaussian form and have "infinite" tails (shown in red in the figure):
...
It is clear that such TS having positive MO=10 points
Once again, to be sure - what you drew, the distribution and the MO taken from the ceiling - is obtained by which variant:
And Seryoga, you must have tested on a large volume your numerous systems based on the Pastuhov's thesis. Can we have a look at their real distribution? I think there should be enough orders there (but of course not a fact).
ADDENDUM:
Just, lots of questions, for example - how did you manage to get a TC-only distribution? What have you got, it's obtained with an infinite deposit on an infinite history????