Market etiquette or good manners in a minefield - page 38

 
paralocus писал(а) >>

It's on the wiener:

You, the first difference from the wiener? - we predict the expected movement. Or did you feed your NS with the original series? And take away the 20, I put it for myself.

Now I drafted code with one linear neuron, it works fine on random process (k=1, d=5):

Look for errors in your indexing.

YDzh wrote >>

I'll just have a look for a minute: from my last graph, where I was going to look for the reason of "miracle", nothing has remained again - iFractals cheated me shamelessly. As soon as I wrote my iFractals it all cleared up, i.e. the miracle disappeared :)

>> Good luck, colleagues.

In fact, it's actually kind of cool! Can you imagine if it had worked out like in that picture... Forex would collapse and we'd all be left without the most interesting hobby. Forever:-)
 
Neutron >> :

Look for indexing errors in yourself.

Nah, I've been feeding it with the original wiener. I'll do the difference now. No more indexing errors, I found them all.


О! This is from the first difference on the Wiener (d=5, k=4), 1000 experiments.

Yesterday I made some attempts to split code, but soon came to a conclusion, that the end-to-end indexing system, shown by you, is the best variant, and the code not scattered in a dozen of different procedures works faster. Now I'm completing the girl to a two-layer

 

Congratulations.

Your girl may only have a straight line in her head, but she's a working girl!

Brag about how you've got this little darling working out predictions of bar sizes from opening prices, e.g. M1. And bring out the cloud for H1...

By the way, the product of the slope tangent on the instrument volatility on the selected TF gives nothing but the profitability of the TS (the average number of points earned, in terms of one bite), subject to entry-exit on each bar.

My profitability for the single linear neuron on H1 for the quid, as a function of the number of entries, is as follows:

Each point is an average of 2000 transactions.

 
I'll do it now. I'll do it in a minute. I never got around to it that time.
 

Look, I have attached my picture above(k=2). You can see a clear excess over the DC commission level. You see, if this is a sustainable effect, you can trade it!

 

Here's what's in the minutes:


Not much to brag about, let's face it

 
Neutron писал(а) >>
... Forex would collapse and we would all be left without the most interesting hobby. For good:-)

Don't worry - there are so many other hobbies... More healthy, in any case :)

 
paralocus писал(а) >>

Not much to brag about, let's face it.

Oh, yeah.

How many entrances and what k?

YDzh wrote >>

Don't worry - there are so many other hobbies... Healthier ones, anyway :)

You're going to take the dummy away from us after all!
 
Neutron >> :
You're going to take the dummy away from us after all!

...rascal -:)

Inputs are now 30 k=2. Tried whitening the entries - seems to be getting better. AUDUSD Minutes


 

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


Reason: