Market etiquette or good manners in a minefield - page 39

 
paralocus писал(а) >>

The inputs are now 30 k=2. Tried to whiten the entries - seems to be better. AUDUSD Minutes

Perhaps not to whiten them, but to even out the distribution of incomes...

And what is the profitability / spread ratio equal to?

Well, it cannot be done on the minutes. Show me on the watch!

I removed the hypertangents from the inputs and outputs and removed the derivative from the error calculation. Here's what I got for 5 inputs, k=2. But it's not minutes any more, it's a clockwork.

Another thing! The question about profitability remained open...

At k=4 everything goes under water.

So, the optimum for a single neuron lies somewhere around k=2.

 

And about volatility - how do you get it? If by ATR, it changes all the time...

One more thing: I slightly warm up the first difference of the series by 10^3 before using it, because figures are too small... is that ok?

 

This is the standard deviation calculated from the opening prices:

The fact that it floats - no big deal, you need to choose n over 1000 (for minutes) and not less than 30 (for hours), then everything will integrate.

Да и еще: я первую разность ряда перед употреблением слегка подогреваю на 10^3 а то там цифры слишком маленькие... это ничего?

It's OK, if you know what you're doing:-) In general, normalize everything to double volatility and we won't have any problems with switching from TF to TF, everything will be normalized automatically. It is very convenient, we don't need to introduce 10^3.

 
paralocus писал(а) >>

...scoundrel -:)

Objection, Your Honour!

We're basically working on the same thing. Just in different ways...

 

Here's the profitability (this is for the picture that was last):





-:) YDzh


 
Is that 0.2 points per transaction?
 

Maybe I counted something wrong? Volatility is correct, tangent too.... Maybe series(X1) is the first difference from Bid on the test sample is wrong?

I will try to increase the number of inputs

 

Doesn't help. It does reduce it to 4, but not by much: 0.223

Aha! Got it! Right.

 

Volatility should probably be counted in pips


 
paralocus писал(а) >>

Probably the volatility should be counted in pips

Probably.

Oh, come on, man.