Shall we play an interesting game? - page 9

New comment
 
LeoV >> :

I've already tested it, in a different software, but the optimal period =14, just like with Vinin. I recommend to optimise the shift as well. It is cooler with period=7 and shift=12.

As I understood the problem is not in search of an optimal period, but in search of something to tie it to, at least to the phases of the moon, respectively, it will be constantly changing. It suggests to take something robust and to put rules of the game on top,

But it contradicts the main useful idea.

 
LeoV писал(а) >>

Well period = 14 the best at shift=1

I don't know what program you use to optimize, maybe there is a wrong system there:), but when solving the problem with a brute force search of periods MT4 shows the best period of 123, the state in the middle of the 2nd page.

 
Figar0 писал(а) >>

I don't know what program you are optimizing in, maybe there is a wrong system there:), but when solving the problem by simple search of periods MT4 shows the best period 123, the state in the middle of the 2nd page.

You know, I'm not going that deep into the parameters to get period=123. Forex market is too fast to take into account 123 last 4 hour bars. Although, to get the maximum profit in this area, maybe this period should be =123, but it's most likely a fit and performance in the future remains under great question.....)))))

 
It's a mystery to me exactly why the machine needs to be optimised, in terms of period. If the task is to create a good auto-optimizer "on the fly", then what is the point of taking the derivative (average)? Take the price (for simplicity - bars) and go! But then we immediately run into the very wall that many deposits and fates have crashed against. No idea here (in this assignment) - no point in fitting. Or am I missing something?
 
Lord_Shadows писал(а) >>
It is a mystery to me why exactly the mask should be optimized, in terms of period. If the task is to create a good auto-optimizer "on the fly", what's the point of taking the derivative (average)? Take the price (for simplicity - bars) and go! But then we immediately run into the very wall that many deposits and fates have crashed against. There's no idea here (in this assignment) - no point in fitting. Or am I missing something?

The point of the task is not to look for inputs, but for critical transient levels. Figaro, on the other hand, sets a problem - we change the period of the mask according to any algorithm. In other words, the result should be a support-resistance level curve that has nothing in common with the classical waving. A simple example is AMA.

 
FION >> :

The point of the task is not to look for inputs, but for critical transient levels. Figaro, on the other hand, sets the problem - we change the period of the mask according to any algorithm. In other words, the result should be a support-resistance level curve that has nothing in common with the classical waving. A simple example is AMA.

I disagree. The aim is to maximise profits for 2008. And as Leonid (LeoV) correctly pointed out, this will have nothing to do with the possibility of getting a self-optimising system based on the average.

 
Lord_Shadows писал(а) >>

I disagree. The aim is to maximise profits for 2008. And as Leonid (LeoV) rightly pointed out, it will have nothing to do with the possibility of getting a self-optimising system based on the average.

Read the thread and Figaro's posts carefully. His interpretation is MA period = any function, not a constant. I.e. essentially adapting the MA period to market changes. You can think of it as a fan of wipers, each one corresponding to a specific market situation. The task of classification.

 
goldtrader писал(а) >>

You saw the point in another way. It's nothing personal. I posted my first result on page 2. It's modest.

I'm not getting much either :(.

So far I've stopped (a lot of manual work when searching for illegitimacy) on the following algorithm

      double dHL = High[1] - Low[1];
      double X1 = 2 * ( dHL) / 0.05 - 1;
      double Y1 = -0.61 + 0.69 * X1 * X1;
      Per = ( Y1 + 1) / 2 * 36 + 2;

So to say GMDH, fuck it :)... (after trying different rsi, ao, etc.)

I'm tempted to play with shifts and other deltas, which aren't allowed in TK!!.. but we understand what the point is ;)

SZZ... When I was a child and computers were so big, I also occasionally came back to the question of searching algorithms of choosing indicator parameters from the point of view of the profit of the whole system...

 
FION >> :

Read the thread and Figaro's posts carefully. His interpretation is that MA period = any function, not a constant. I.e. essentially adapting the period of the MA to market changes. You can think of it as a fan of wipers, each one corresponding to a specific market situation. The task of classification.

Sergey, I understand that the waving can change... I also understand that we can teach it, so to speak, to understand the market situation... I do not understand how to get the maximum profit in 2008, you can earn in 2009, that's one. And why then not to look for auto optimizer for bars, it will be sharper, more accurate and flexible system, that's two.

And about the waving fan is the very first idea that comes to mind...

 

Attempts depending on the day of the week:

Net profit: 11260

Profitability: 3.64

Deals: 251

Optimal values:

Interestingly, the optimal period rises linearly from Monday to Wednesday (86,126,166) and then falls sharply on Thursday and even lower on Friday

You can consider not just the day of the week, but also depending on the 4 o'clock. The optimal periods look something like this

on the abscissa axis, the hour of the day from Monday at 0:00 to Friday at 20:00.

Profit rises to about 18t and PF>4, but this is a clear memory

1234567891011
New comment