Unprofitable trades 0!!!!!! - page 5

 
Hoper23 писал(а) >>
Well, what if the optimization is carried out on one period, but when tested on other dates it shows the same result?

This is Out Of Sample - one of the methods of testing a trading system for stability.

Hoper23 wrote >>

You mentioned the drawdown - how do I reduce it?

I don't know... it's a question of how to trade profitably. Everyone solves it in his own way. Of course, I will suggest my own methods of drawdown reduction. But I must warn you, I was criticized for them more than once. They said that this approach leads to system overoptimization. And I think so. If these methods work for me, it means that they are suitable for me.

1. I consider it an axiom that forex is asymmetrical relative to rise/fall. Therefore, I optimize some of my systems (NOT ALL, but only some of them) separately for buying and selling. That is essentially I make two systems from one: one only buys, and one only sells.

2. this method follows from the first. Its essence is the differing sizes of lots for buying and selling.

3. I check trades for interdependence. This idea is taken from the book "The Mathematics of Money Management" by Ralph Vince. I calculate two indicators: Z-score and correlation. Using methods of lot size management in two versions: after the previous losing trade and after the previous profitable trade.

 
KimIV писал(а) >>

1. I consider it an axiom that forex is asymmetrical with respect to rise/fall. Therefore, I optimise some of my systems (NOT ALL of them, but only some of them) separately for buying and selling. That is, I essentially make two systems out of one: one only buys, and the other only sells.

Is there a criterion that can show qualitatively or quantitatively the correctness of this statement (for now)?

 
Neutron писал(а) >>

Is there a criterion that can quantitatively or qualitatively show the validity of the statement (so far)?

Dependencies between days of the week

Of course, this is not enough for a rigorous proof. We would need to consider the dependencies between bars of several more timeframes. We would also need to make statistical studies of the correlations:

- price change down / how many bars it occurred in = speed of fall

- Price change up / how many bars it happened = growth rate

Further, if desired, you could study the acceleration (Momentum indicator).

But I don't think it's necessary. For me it was enough for formation of my market model. If somebody cannot do it enough, let him dig deeper.

 
KimIV >> :

I optimise separately for purchases and sales. In essence, I make two systems out of one: one that only buys and one that only sells.

It's a very interesting method, there's a real logic to it. I hadn't even thought about it before. I will now try it for myself...

 
KimIV >> :

I calculate two indicators: Z-score and correlation. I apply lot size management methods in two variants: after a previous losing trade and after a previous profitable trade.

How do you calculate these 2 values? And how do you manage lot size in + - trade? I understand the volume, but which one goes where...

 
KimIV >> :

I calculate two indicators: Z-score and correlation. I apply lot size management methods in two variants: after a previous losing trade and after a previous profitable trade.

How are these two indicators calculated? And how is lot size management performed in + - trade? I understood by volume, but which one...

 
KimIV >>:.. I consider it an axiom that forex is asymmetrical with respect to rise/fall...
Am I correct in assuming that it is a constant asymmetry, independent of the current trend?
 

to KimIV

Thank you. I see what you mean.

Another interesting thing is this. If we take a sufficiently long time interval (the number of bars > 1000), we can ascertain that the number of positive increments tends to the number of negative ones. On the other hand, the average absolute value of increments (on the selected TF) depends on the instrument price, i.e. <dS/S>=const. But then with the equality of upward and downward movements we are bound to have different amplitudes of these movements - the average amplitude of the upward movement will always be a bit larger than for the downward movement...

Is there any way to exploit this?

 
Shit, split, but the result on drawdown is the same - with maximum output comes maximum drawdown.
 
Hoper23 писал(а) >>

And how are these 2 indicators derived?

Google it yourself... formulas are not military, easy to find...

Hoper23 wrote >>.

And how do you manage the lot in + - trade? I understood the volume, but which one goes where...

This is different for every system. This is also an optimization problem. I use two criteria:

- minimum drawdown,

- maximum recovery factor.

As an example: after a loss lot 0.1, after a profit lot 0.3. This is for systems with a negative Z-account.