Market etiquette or good manners in a minefield

Neutron
2825

Hooray! - My three-layer non-linear neural network has started to show stable positive trading results. And immediately the question about the optimal MM was raised. I already touched this subject here before but considered a case without taking into account brokerage companies' commission (Spread). I'll try to obtain the basic formula for the optimal size of an open position as a function of reliability of TC prediction (1/2+p, where p is the probability of correct prediction of the expected price movement sign), and for the optimal size of the average take (profit) module <|S|>.

Let it still be:

S - the price of the instrument in pips,

dS - price increment for the time of holding the position in points,

K - deposit size in $,

dK - deposit increment in $ for the time of position holding,

stLot - price in $ of the standard lot,

Lot - the number of lots to be opened,

Lever - trade leverage.

That seems to be all. Let's write down the basic equations.

stLot *Lot= K*Lever - connects size of an opened position (Lot) with the trading leverage and the size of the deposit.

dK/K=Lever*dS/S - relation of the relative value of the deposit increment to the relative value of the exchange rate fluctuation.

Then the basic balance equation ( see details here) including the spread would look as follows:

This equation shows the deposit size after n transactions, taking into consideration all values we are interested in, and first of all the accuracy of the TS forecast for the sign of the expected price movement p.

Let's find the optimal value of the average size of bribes |dS| and the value of trading leverage Lever in the standard way:

We can see that there is an optimal size of an open position, which depends on the deposit size, prediction accuracy and DC commission. Both its exaggeration and understatement underestimate the possible profit rate. In addition, the average size of TC (in points) should also be optimal and determined by the commission and the accuracy of forecasts.

Taking into account smallness of the parameter p (<0.1), one can simplify expressions:

Here expressions are written through value characterizing accuracy of prediction p. The third equation estimates characteristic time of deposit doubling. We can see that it depends very much on the forecast accuracy (as the fourth degree from the precision and the second from the commission). Thus, the main attention a trader should pay is to improve accuracy of TS forecasts.

Hide
2572
Hide  
the link to "When, the basic balance equation (details here) including the spread would look like this:" does not lead anywhere
suvorov
12
suvorov  

Where did you get these formulas from anyway? Did you even understand what you wrote?

Given the smallness of the p parameter (<0.1), you can simplify the expressions:


With such a probability less than 0.1 it is better not to trade at all:))) it will be more expensive

suvorov
12
suvorov  
Another question is how to determine the probability of predicting the correct sign of the expected price movement? Statistics? Ok, for example: TS gave the signal to open a long position the first 2 hours the market goes in the right direction, and 2 hours later went down. So the TS seems to have given the right result, but it seems not.
Hide
2572
Hide  

It says "(1/2+p, where p is the probability of correctly predicting the sign of the expected price movement)", i.e., p should be added to 0.5


Why the top-starter needs such difficulties, is not clear yet.

Prival
4981
Prival  
Neutron писал(а) >>

....

Good +10. Finally someone has expressed it in a formula. There is 1 more fine-tuning to be done in the formula. Confidence interval of prediction probability, because this value can only be estimated from history, and since it is an estimate, it should have a confidence interval.

There could be a good article if you team up with Mathemat and its bernuli. Something like "How to correctly calculate Lot by looking at the tester report".

Sergey Fionin
1454
Sergey Fionin  
Neutron писал(а) >>

Hooray! - My three-layer non-linear neural network has started to show stable positive trading results.

Knowing your love of analytical methods I am interested - what do you feed to the output of the network for training?

Дональд
772
Дональд  
Neutron писал(а) >>

Hooray! - My three-layer non-linear neural network has started to show stable positive trading results. And immediately the question about the optimal MM came up. I already touched upon this subject here before, but considered the case of

Of course I'm sorry, but being not a fan of "magic ointments" which neuro-adapters are, I want to say that I cannot remember when an EA I wrote did not show stable profit on the whole history without optimization and so on. So I don't understand what the problem is, for fuck's sake? - Create a profitable EA? Why all this fuss with neural networks?

But I hasten to answer the detractors of "grails" - in addition to the fact that the "grail" must be profitable it should earn the "right" amount of money, and not 1000 quid for 10000 ... per month. Because 100000 quid it will not make any more, and less than that is not even worth bothering .... IMHO ... :)))

Sorry.

Sergey Fionin
1454
Sergey Fionin  
NProgrammer писал(а) >>

Why all the fiddling with neural networks?

In your neighbour's thread about "sharing etc" you have real, uncomprehending fucks with the "right mashka", and by all appearances all your advisers are very profitable and you are a millionaire, but you come here to make a fool of yourself.

Дональд
772
Дональд  
FION писал(а) >>

In your neighbour's thread about "sharing etc" you have real, uncomprehending fucks with the "right mashka", and by all appearances all your advisers are very profitable and you are a millionaire, but you come here to make a fool of yourself.

I do not come here to educate anyone, 100%...

I'm telling you a very clear understanding of "what I'm talking about"... Forex is so complicated that ten neurons cannot do it... Not in a million years. If you like, it is more complicated than chess, and in general it is meaningless chattering, but how many neurons does a frog have? And how many you have. Have you ever tried to use at least one thousandth of your neurons and learn to trade manually? And when you learn, I suggest you get back to neural networks and I assure you, you will react in the same way to all these "magic ointments" - you smeared on them and became irresistible, it's like paganism ...

Sorry about that, I won't say another word.

Sergey Fionin
1454
Sergey Fionin  
NProgrammer писал(а) >>

I come here for sure not to pre-educate anyone, 100%...

I'm telling you very clearly I know "what I'm talking about" forex is so complex that a dozen neurons can't handle it... Not in a million years. If you like, it is more complicated than chess, and in general it is meaningless chattering, but how many neurons does a frog have? And how many you have. Have you ever tried to use at least one thousandth of your neurons and learn to trade manually? And when you learn, I suggest you get back to neural networks and I assure you, you will react the same way to all these "magic ointments" - you smeared on them and became irresistible, it's like paganism ...

Sorry about that, I won't say another word.

What can I say? Everybody probably thinks they know "what they're talking about"...