Whats your best method to follow historical price peaks?

 
What I mean is counting peaks. I know zig zag is widely used for this. But I find the depth setting a little too inaccurate. They can find good peaks but they also miss peaks even on the lowest settings. Does anyone know a better way? I have a very good trading strategy I want to code, but it depends on getting this part of it accurate. This is the hardest part of the coding.

Also I know fractals are used a lot for price peaks but they give too many signals and lag too much. Well with this method I am trying to program the lag doesn't really matter, but the number of signals does... I find the zig zag the best method that I know of, but I am hoping to find smarter people here who can help me out.


 
sgsummers:
What I mean is counting peaks. I know zig zag is widely used for this. But I find the depth setting a little too inaccurate. They can find good peaks but they also miss peaks even on the lowest settings. Does anyone know a better way? I have a very good trading strategy I want to code, but it depends on getting this part of it accurate. This is the hardest part of the coding.

Also I know fractals are used a lot for price peaks but they give too many signals and lag too much. Well with this method I am trying to program the lag doesn't really matter, but the number of signals does... I find the zig zag the best method that I know of, but I am hoping to find smarter people here who can help me out.


You simply calculate the annual standard deviation from the mean value of the statistical normal distribution. Anything above/below that with a certain factor is a high or low point. 
 
Thank you for the reply. I am not sure I fully understand. Would you be able to post an example?
 
tomm73:
You simply calculate the annual standard deviation from the mean value of the statistical normal distribution. Anything above/below that with a certain factor is a high or low point. 

This is the same as D1 Bollinger Bands over as many periods as annual trading days. I don't think it helps because the question was about individual extreme points, whereas with standard deviations price could stay at extrema for dozens of consecutive periods (=regions rather than individual price points). Therefore fractals or ZigZag are probably already the better idea.

@sgsummers: if a point actually is extreme (=peaks / troughs) is defined relative to its surroundings and with price half of the "surroundings" (=price neighbors) always are in the future, which is why I think it's by definition that you can evaluate those extreme points only in hindsight, so that "lag" seems obligatory with any kind of extreme point evaluation.

In this context peaks/troughs with a higher significance need more neighbors in order to be defined or - in other words - to verify that they are really standing out. Therefore high significance and low lag is a contradiction in itself.

 

Similar   to  zig zag you can count extreme peaks  and troughs  with SSA indicators  but only in hindsight : https://www.mql5.com/en/forum/174385/page321#comment_4564388  and  https://www.mql5.com/en/code/7359

How do  you  count and compare ?Using  Spearman's correlation ?

Coding help
Coding help
  • 2014.08.02
  • www.mql5.com
Hi, Mr. Guru could you please help me to constantly refresh an ea after each tick movement...
 

Thank you guys for all the info. I will try to look it over in detail soon. Hindsight is just fine. That's what I am looking for. I don't care to predict or confirm current peak. I want the most accurate way to follow the historical peaks mechanically. I have my own way to predict the future peaks but I need to get this part first. I can do it by eye but I want an indicator.

And I am not necessarily looking for the most extreme peaks. Just peaks. Actually my method would project the extreme peak off of normal peaks. So I'm really just looking to trace the market historically but more accurate than the zig zag.

Reason: