What to feed to the input of the neural network? Your ideas... - page 17

 
Ivan Butko #:

"Humbling is usually started from intellectual impotence."


Victor Grigorievich is just a maestro.

Great post, though.

 
Aleksei Stepanenko #:

Can you explain in more detail. A deal has an opening point with two coordinates and a closing point with the other two coordinates. You look at the chart and think about what? What's the probability that the trade will be profitable. Right? What information gives you the answer to that question?

history, but there is a catch: what interval to take (1 year, 2 years or 5 months) is the question! I'm not kidding, this is the most important question. For now I consider 1 year, but maybe other developers will choose a different interval.

 
Sergey Pavlov #:

story, but there's a catch: what interval to take (1 year, 2 years or 5 months) is the question of questions! I'm not kidding, this is the most important question. For now I'm thinking 1 year, but maybe other developers will choose a different interval.

why a year? for statistical reliability of the number of trades? probably depends on the TF?
 
Andrey Dik #:
Why a year? For statistical reliability of the number of trades? I guess it depends on the TF?

A year for M1 minutiae

 
Sergey Pavlov #:

year for M1 minutes

Wow.

Is it brute force learning in the MT optimiser or real through weight training?

 
I don't see the point in EAs that trade 1 trade per month, you should trade as often as the neural network allows.
 
mytarmailS #:

Still a great post though

I agree, mental hygiene

 
Ivan Butko:
Tried to feed in:
-closing prices
- difference of closing prices of N candles in a row
- difference of closing prices of N candles in a row from all allied pairs on both the euro and dollar sides, on the eurodollar pair
- ratios of candles' shadows to their bodies of N candles in a row
- ratio of closing price to all OHLC of all N candles in a row
- all of the above with N and N*k gaps between candles


And....I've run out of ideas))))
What to feed to the input of the neural network? Your ideas ...

What's stopping you from making a big automatic search of many options?

I think this is the best option, experimentation is the criterion of truth, if you make the experimentation automatic, then it's a fairy tale.

 
Ivan Butko #:

Wow.

Is brute-force learning in the MT optimiser or real through weight training?

MT tester will not work, create your own tester.

 
Sergey Pavlov #:

The MT tester won't work, you have to create your own.

Yeah, so I'm a long way from your car.

But I'll try my own method, that's good.