is there any normalized indicator without divergence?

 

hi guys, sorry for my bad english, it is not my native language,

i need an indicator which has no divergence like stoch, rsi or macd, it should show the real picture what price has done, without divergences?

if it is already in a fixed range it would be great, any suggestions?

thanks ;)

 
let me explain, in the picture you see a stochastic(200,1,1) and a rate of change oscillator(200) (in this approach i try to normalize the roc with (val-mean) / std)

i'm unable to get a normalized UNCHANGED view on price.



the blue rectangle shows that stoch is and stays < 20, but price is falling down more and more, i don't want this.

the red trend lines are showing a divergence in roc, i also don't want this.

all i need is a way to get the real movement in a fixed range?
 
The only indicator I know of, that has no divergence and indicates perfectly what the market is doing; is called "Price". It is the best indicator you will ever find!
 
nina32: i need an indicator which has no divergence like stoch, rsi or macd, it should show the real picture what price has done, without divergences?
  1. Do you mean divergence or no repaint?
  2. Indicators do not show what price has done, price shows what price has done.
 
@FMIC

"It is the best indicator you will ever find!"

That is true, i also know this, but i'm unable to get the Price in a fixed range, because it can make new extremes all the time.

@WHRoeder

1. I got no repainting issues, why you think that?
2. Thanks but that does not answer my question?

To explain a little bit more, i'm using a neural net (fann/fann2mql). If i for example use the rsi and train my NN there are situations were the rsi is falling but the price does not (or vice versa), a so called divergence.

If i use the rsi as an input to the NN it "can not really see" what the price is doing, rsi tells the NN that we are going down - but if price doesn't do that - my net gets confused.

I also know that we can normalize a value (price/close) with 2 ways, normalizing by history high/low extremes or with mean & standard dev (val-mean) / std.

The first approach (normalizing with extremes) has the problem that older data get squeezed if new extremes appear, everything the NN has learned in history isn't correct anymore.

The second approach (normalizing with standard dev and mean) has the problem that outliers (new extremes) are filtered if they are > STD, it also removes important information's.

In short, all i need is the unchanged price movement in a fixed range - 1 <> 1, sounds simple but i can't find the right way...
Reason: