Subsystem "Asset Management" - page 6
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
I'm going to give you a brief summary of what I've had time to do at the moment. I will tell you briefly about what I have managed to do at the moment of "now". Preliminary conclusions: Markov chains are not suitable for this problem, but I can try them for one tool. Linear programming works - "but how it does it..." (С). :о)
In the first iteration to use LP I have simplified the model even more (while I learn :o): Instead of getting lots directly from the LP model I used a smoothly varying deposit value, assuming that later I can easily calculate the lot, took the cost of one pip as a constant, didn't consider spreads at all. In the first iteration I gave up risk limits and considered the time of existence of a segment.
For calculations I used the following software (the example for three variables and five limitations is shown - actually these are models and all I need to "push" into such a thing):
http://www.mathelp.spb.ru/applet/SimplexTool.htm(theory is also there)
Of course there is MathCAD, MathLAB, Exel and a lot of other stuff, but it's more than enough for beginners. Further, if you enter the notations:
p - Difference between opening and closing price - (TR triggered)
x - unknown amount of deposit to make a trade operation
then I took the target function as follows:
Limitations of the model were taken as follows
-limitations on position holding (value of SL)
-limitations on the maximum lot size
-Restrictions on the total deposit size (the sum of all "parts" of the current deposit per each symbol should not exceed the total deposit size)
As expected - it works, i.e. it considers the optimal lot size. But, - but if SL is put within reasonable limits, it is possible to get the optimal size for small amount of unknowns by simple operations and using "if" operator. It makes no sense to use LM for such a simple model.
BUT - but the next iteration is already the inclusion of the duration of the transaction and risks. With duration the idea is simple. The markets are very dynamic and it's not safe to "sit" for a long time with open positions (especially at ) and therefore it's desirable to include duration estimation, e.g. in this form - to give preference to zigzag segments with minimal values of t. Here is the most interesting thing: if we introduce an empirical dependence, say, exponential, we actually get the formula: introduce the time - we obtain the "optimal" value for the instrument. As many traded instruments - so many "optimal" values of the deposit for trades obtained without taking into account the trading environment and the current situation. Then the task changes a little - in fact, it is a choice. And with risks in general a separate subject, while I think over the formula "the less I risk, the more I earn", it is somehow closer to my emotional equilibrium.
PS: this is just the beginning, I think there will be a sequel soon.
to eire
А что уважаемое сообщество думает о книге Ральфа Винса "Новый подход к управлению капиталом.
Asset Allocation Structure between the Different Investment Instruments"?
I have to read it. Where can I find it?
to thecore, Neutron
I hope to get to the lab soon and take a close look. :о))))
to anubis
Similarly - I'm also studying algorithms.
what about the restricted section? can regular mortals get in there ? =)
it's a private club, otherwise the whole point is lost :o) Just kidding - I'm invited myself.
Linear programming :-) it's as simple as five kopecks.
Conclusions:
Prival for the Schnobel Prize - get in line after me. And don't forget to write "2" on the palm of your hand, so you can control the queue. But seriously, Prival - add to your condition the consideration that the price is not a constant for all points, take into account the duration of transaction, take into account the risks (which is a whole theory), take into account that the end of the waves do not always coincide, a full-fledged accounting for position retention. And try again to multiply something by something.
PS: Prival, I just do not understand - do I really have such a bad "articulation" and you, dear - do not understand what I write? I've written clearly - I'm just studying LP, made a couple of simple models to understand how it all works. At the next iteration I will complicate the model, add a bunch of stuff to it. In the first post I drew a diagram of my system, marked what I want to discuss, if I have someone to talk to. And now, do you really think that your "old" and "good" opponent is dumber than a locomotive and does not understand that the accuracy of the prediction is important?
.... But seriously, Prival - add to your condition the consideration that the value is not a constant for all items, take into account the duration of the transaction, take into account the risks (and this is a whole theory), take into account the fact that the end of the waves does not always coincide, a complete accounting for withholding positions. And try again to multiply anything by anything....
All this just comes out if there is no forecast. There is a theory of STRM (statistical theory of optimal control), and so there is nothing without the prediction or the so-called endpoint of control.
I try to be a "good" opponent :-)
to eire
I have to read it. Where can I find it?
http://nehudlit.ru/books/detail7024.html
This is exactly what happens if there is no prediction. There is a theory of STOU (statistical theory of optimal control), and there is nothing without the prediction or the so-called endpoint of control.
I try to be a "good" opponent :-)
Prival, yes, there's a lot of stuff, ... that can come out .... :о)))
http://nehudlit.ru/books/detail7024.html
>> Senx, let's read it...
Isn't it?
I don't think it is.
I'll explain why.
1. We take different instruments into the portfolio. Currency instruments are chaotic in nature. Sometimes they are chaotic in nature.
uncorrelated sometimes highly correlated. So you cannot take completely uncorrelated instruments.
2) Based on the previous point, it's obvious that it's impossible to develop absolutely identical strategies for different instruments.
for different instruments.
3. If the selection of instruments and trading strategies is unsuccessful, it is possible to obtain the result,
when due to correlation of instruments the strategies will have a bigger drawdown than for one instrument in particular.
4. The deposit is not infinite, so we cannot talk about 100 or even 10 instruments to work with.
These are the simplest arguments.
The task of a trader in such situation is not to provide maximum profit, but not to allow the drawdown below the critical level.
To do this, we need to study strategies for the selected instruments in order to determine the maximal drawdowns
we can expect. And if at any moment the maximal drawdown exceeds the one that we have got as a result of the study
history we should stop trading and either revise our strategies and portfolio or wait
or wait for the market to enter the desired direction for the selected strategies.
Accordingly, your assumption can be valid only in a limited segment of the history,
and it may not work at the other one. And we don't know in advance which sector will be ahead.
It seems to me that the state of stock indices like S&P500 confirms my assumption.
There too, an uncorrelated portfolio and strategies are picked up, but when the economic
The crisis collapses (becomes highly correlated).
p - Difference between opening and closing price - (TR triggered)
x - unknown amount of deposit to make a trade
I took the target function as follows
SUM {x*p}->max
Sergey, what is the dimension of your target function, dollars squared? You may use relative values - it's more correct, otherwise you may get these square quid from a brokerage company and then -)
to thecore
It seems to me that not so.
When I chose a model, I was primarily guided by its simplicity and clarity, trying to convey the main idea through its logical transparency and consistency, which is sufficient for the level of communication on the forum. In reality, as you have correctly noted, more complex structures need to be applied, including correlation relationships between instruments in the portfolio and TS, as well as features of bribe distribution functions for each TS (they determine the possible risks). This and much more, constitute the scope of the problem that an investor needs to solve when compiling the optimal portfolio. The exact solution in this formulation is difficult, we can only approach it using some or other assumptions. This is creativity multiplied by knowledge. No wonder I was surprised at the immensity of the problem posed by the author of the topic - when solved in full, it is really a work worthy of public recognition.
So, it does not seem to me that something is wrong.
Sergei, what is the dimension of your target function, dollars squared? You should use relative values - it's more correct, otherwise you'll get these square quid at the brokerage company cash desks and what to do with them :-)
Seryoga, what is it with you? It's all right! It's obvious that you haven't traded for a long time :o)))))))))) You probably only work in MathCad :o)))) This is how the profit is calculated:
Profit/Loss = (Contract * ClosePrice) - (Contract * OpenPrice)
p this is ClosePrice-OpenPrice- (the difference between the opening price and the closing price - (TR triggered)), added - ugh - reverse, "well you get me men" (C) :o)))
x is Contract ( - unknown amount of deposit for trade operation)
More details here: http://www.alpari.ru/ru/help/168.html
PS: you can count from the cost of the item, but its value varies over time, not very convenient. As a matter of fact I have already made such a model in LP, taking into account risks. But it's too early to post it :o)