same shape between price and your indicator?

New comment
 

This is not programming related, so please humor me on this one.

Don't you find it annoying that price is going up when your indicator is going down? I used to think that the best way to counter that was finding some magical adaptive window function (that changes the look-back length of your indicator dynamically). So I have looked into that, but I never have found the right solution.

How about we do some brainstorming here? It doesn't always have to be forex-tsd right?...

I looked at a number of windowing functions; would anybody be so generous enough to give me a hint?

Like this isn't this frustrating:

 
If indicators were a mirror image of the price, for what reasons would we need indicators? Price have no dependence on the previous bars however indicators do. Price have no set look-back periods however indicators do. Price have no up-side limitation and no 0-line to oscillate. I dunno, but to me this is like asking the moon to look like the sun; instead of using the moon to locate the direction of the sun.
 
ubzen:
If indicators were a mirror image of the price, for what reasons would we need indicators? Price have no dependence on the previous bars however indicators do. Price have no set look-back periods however indicators do. Price have no up-side limitation and no 0-line to oscillate. I dunno, but to me this is like asking the moon to look like the sun; instead of using the moon to locate the direction of the sun.


So I guess you recommend price action strategies? Which one in particular. I have been looking at a pinbar strategy, but not having much success. In fact after 3-4 months of hard toil as a relative newbie at automated trading, the best I was able to achieve is a nice curve fitted MACD strategy that obviously failed in forward testing.

Also, how does price action differ from an indicator? Would the price action of H1 not be the same as 12 periods of M5? Would an 11 period indicator on M5 starting at 1 not give a better approximation of "price action" with only 5 minutes lag and not 1 hour?

 
rocketman99:


So I guess you recommend price action strategies? Which one in particular. I have been looking at a pinbar strategy, but not having much success. In fact after 3-4 months of hard toil as a relative newbie at automated trading, the best I was able to achieve is a nice curve fitted MACD strategy that obviously failed in forward testing.

Also, how does price action differ from an indicator? Would the price action of H1 not be the same as 12 periods of M5? Would an 11 period indicator on M5 starting at 1 not give a better approximation of "price action" with only 5 minutes lag and not 1 hour?

Wasn't recommending anything. I was giving my opinion about why Indicators shouldn't be an exact relative value of price. I suggest you put more work into the money-management, instead of looking for more accurate signals. For Macd, I had the exact same experience within the exact same timespan. Don't use strategy optimizer, don't go parameter hunting if the strategy isn't already profitable on first logic.

Also, how does price action differ from an indicator? Given this topic, Indicators don't have the same values as price.

Would the price action of H1 not be the same as 12 periods of M5? I don't like the term Price-Action, matter of fact I Hate it. But Answer=No. and No.

 
ubzen:

Wasn't recommending anything. I was giving my opinion about why Indicators shouldn't be an exact relative value of price. I suggest you put more work into the money-management, instead of looking for more accurate signals. For Macd, I had the exact same experience within the exact same timespan. Don't use strategy optimizer, don't go parameter hunting if the strategy isn't already profitable on first logic.

Also, how does price action differ from an indicator? Given this topic, Indicators don't have the same values as price.

Would the price action of H1 not be the same as 12 periods of M5? I don't like the term Price-Action, matter of fact I Hate it. But Answer=No. and No.


Ubzen, having a bad day? Your tone is rather aggressive in this thread and you are not really clarifying much.

Anyway, still don't understand the following:

> Price have no dependence on the previous bars however indicators do

To me an H1 bar is still an "average" of one hour? Do I get any real insight of market direction in that single 1 hour bar? Not really. I need to look at things in context related to other surrounding bars which now all of a sudden sounds a lot like an "indicator" to me based on past history? Does that single 1 hour bar in isolation tell me if most of the bar was spent at the top or bottom of the 1 hour range?

> Would the price action of H1 not be the same as 12 periods of M5? I don't like the term Price-Action, matter of fact I Hate it. But Answer=No. and No.

Explain.

Sorry but I do not fully understand the difference between "price action" and indicators and would really like somebody to explain. To me price action is just another way of reading the tea leaves based on past history.

 
if you study each indicator or oscillator, how it's built & what calculation has been used to build it & for what purpose it's been made, you will find your answer
 
ubzen:
If indicators were a mirror image of the price, for what reasons would we need indicators? Price have no dependence on the previous bars however indicators do. Price have no set look-back periods however indicators do. Price have no up-side limitation and no 0-line to oscillate. I dunno, but to me this is like asking the moon to look like the sun; instead of using the moon to locate the direction of the sun.

"If indicators were a mirror image of the price, for what reasons would we need indicators?"

-> To remove the noise and clearly show the trend in real time, or as much as possible.

"Price have no dependence on the previous bars however indicators do."

-> Calculate auto-correlation and partial acf: you will see something else. Or mutual information: same story.

"Price have no set look-back periods however indicators do."

-> Sounds kind of obvious but actually I think it is a valid point. But what does price have that can act like a reference?

 
qjol:
if you study each indicator or oscillator, how it's built & what calculation has been used to build it & for what purpose it's been made, you will find your answer


Yes Master Po.

:)

 

@rocketman99: Ubzen, having a bad day? I'm having a good day actually :). Notice that I didn't recommend looking at One-1-Hour_Bar or any forecasting method for that matter. You're still talking about forecasting/signals; this is one of 3-legs which holds up the table.

What I meant was that the H1 value was not created by considering the values of the bars which preceded it. Most indicator/oscillator uses averaging to minimize noise. The Indicators existence is solely to help indicate the probable Next-direction of price. In isolation most single-study ie: indicator or price-action cannot be used alone but rather combined to create a system.

Like I said, I don't like the term price-action. Price-Action is a broad term which implies studies outside of indicators. I rather like to call them for what they are. For example: a Pin-Bar cannot be found on One-1-Hour_Bar because one needs at least 3-Bars last time I checked. But a Pin-Bar can be found within Twelve-5-Minutes_Bars. A doji however can be found on One-1-Hour_Bar. The value of the price move within 1Hour or 12_5Minutes didn't change but their So-Called price action did.

 
MrH:

"If indicators were a mirror image of the price, for what reasons would we need indicators?"

-> To remove the noise and clearly show the trend in real time, or as much as possible.

"Price have no dependence on the previous bars however indicators do."

-> Calculate auto-correlation and partial acf: you will see something else. Or mutual information: same story.

"Price have no set look-back periods however indicators do."

-> Sounds kind of obvious but actually I think it is a valid point. But what does price have that can act like a reference?


-> Then IMO, Price itself is the only indicator for this task. For your same shape between price and your indicator, All I can seem to imagine is a Line-Chart(of price) within an Indicator window. You can change the relative Scale all you want but the curvature will remain the same. How do you intend to turn this into a prediction?

-> Looking at Correlation of 1, is the same as looking it Price-itself. Looking at Partial ACF is like looking at fractions of the same correlation, it's like relative Scale. AE once said Can't do the same thing, and expect different results ;-)

-> Believe me when I say I do not know the answer to that question.

 

Indeed oszillators tend to give fail signals after extreme points.

some thoughts:

  1. maybe this 'unwanted' behaviour can be used as a indicator to trade after such extreme points
  2. by looking at the base of the oszillatators you can identify the 'unwanted' instantly
  3. The two indicators you have posted are quite different
  1. The first one is some kind of force indicator with a moving average, the result is not unexpected at all.
  2. The second one is more interesting, the two lines could be used as trendfilter while the distance between these two can be used as force
  • If you want to eliminate the unwated issue you need to take the base into your calculations.

Example:

Take the MACD:

If you compare the result with the base you are quickly able to spot the bars where the values do not follow the pricemovement. (green bars on the second MACD)

Whats left is some calculation to adjust the MACD in such circumstances

      double imaF=iMA(NULL,0,FastEMA,0,MODE_EMA,PRICE_CLOSE,i);
      double imaS=iMA(NULL,0,SlowEMA,0,MODE_EMA,PRICE_CLOSE,i);
      MACDS[i]=imaS;
      MacdBuffer[i]=imaF-imaS;
      if((MacdBuffer[i]>MacdBuffer[i+1] && MACDS[i]<MACDS[i+1])){
         MACDSTRUE[i]=MacdBuffer[i]-(MathAbs(MacdBuffer[i+1]+SignalBuffer[i+1]));
      }
123
New comment