What to feed to the input of the neural network? Your ideas... - page 51
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
...
Anyway, there's more food for thought about the angle.
In my opinion, it is not enough to set the distance in price values between bars
In my opinion, it is not enough to set the distance in price values between bars
The problem of scaling when calculating the angle has already been discussed on the forum, so we can state that you can play with the formula, depending on what mobility and what values the user wants to see in the output.
I printed, it really showed 40, 63, 77, -34, -55 degrees. Basically, this is already what is needed from the angle.
...
...
...
Anyway, there's more food for thought about the angle.
Yes, you can not limit yourself to a strict window of N bars.
You can set the angle on the extrema of ZigZag or the nearest fractals. And for each bar, feed the angle between the current price and these extrema. Any extremum is a priori more important than its internal prices. After all, it indicates the redemption of positions (reversals and the like).
Yes, you can not limit yourself to a strict window of N bars.
You can put the angle on the extrema of the ZigZag or the nearest fractals. And for each bar, feed the angle between the current price and these extrema. Any extremum is a priori more important than its internal prices. After all, it indicates the redemption of positions (reversals and the like).
What angles?
Wake up...to talk about "angles and degrees", the dimensions (metrics) of the axes must be the same.
What kind of angles?
Wake up...to talk about "angles and degrees", the dimensions (metrics) of the axes must be the same.
Limited Thinking
There's a formula, there's two perpendicular sides. Go ahead and with a song.
You convert one of the sides to a real number, or the other side to an integer and voila - it shows you the real degree.
The fact that it can vary is a matter of mobility. The main thing is that the conversion factor should be all over the graph.
Mashstability.
UPD
Distortion - does not matter if it is all over the graph.
Candlestick formations are almost the first thing a neural network checks))) After increments.
Here, unfortunately, personally on my experience - all candlestick formations "head-on" wagering on the forward 50/50 up and down. With a margin of error in the form of a long-term trend. Then the shift will be a couple of per cent and it will be 52 vs. 48.
But, the trend doesn't have to continue, so the rules "The trend is your friend" and "The longer the trend, the more likely it is to reverse now" are two equally useless opposite theses.
This is why I made the point earlier about context. That is, when the same candlestick formation is traded today as a buy and tomorrow as a sell.
In this sense, neural networks need to analyse the market context and decide when a specific candlestick pattern appears - where to trade it now - up, down - or abstain.
This idea is similar to filtering with the help of several indicators, as some people do. Only there the trade goes in one direction anyway. And in our case it should be both ways.
Yes, you can not limit yourself to a strict window of N bars.
You can put the angle on the extrema of the ZigZag or the nearest fractals. And for each bar, feed the angle between the current price and these extrema. Any extremum is a priori more important than its internal prices. After all, it indicates the redemption of positions (reversals and the like).
z = soft+warm/2;
z = soft+warm/2;
For the especially gifted:
If there is a visual embodiment in the form of 2D graphics, it means the presence of geometric objects in it.
And as you do not call scales X and Y, crocodiles and hippos, soldiers and refrigerators, communists and crossword puzzles, the main thing - numbers. Divisions. E-d-i-n-i-n-i-ts-y.
If one side on the X scale has 10 units, and the perpendicular side on the Y scale also has 10 units, then inside this 2D graph we can construct a RIGHT Angle and calculate its angle.
I managed to figure out how to feed the angle to the input
If you look at the picture, then yes, warm and soft, but in the code everything is normal.
The brackets in the picture are wrong. It should be like this.