Possible to code this 3D heatmap in PRT? (Mark Whistler’s Wave PM)

• 06/26/2019 at 6:50 AM #101418
  jebus89

  Hi so the indicator im talking about is found here: https://www.tradingview.com/chart/BTCUSD/6SIl2X3f-The-Holy-Grail-of-Trading-Advanced-Volatility-Theory/

   

  Its the 3d heatmap indicator which is explained like this:

   

  “So you may be thinking to yourself, would it be possible to measure the longest period of mass to find the longest possible pivot that will define trend and mean reversion? Why yes it is. I’ve had my developer create a 3D heat map representing 32 separate periods of mass. Red represents a mass level above 0.9. This allows you to define which periods are overexpanded, your longest overexpanded period, and volatility overhead (longer period distributions compressing that will define lateral trading as represented by their 1.25 stdev.) ”

  Scroll down until you see the indicator.

  Thanks.

  

  jebus89

  Participant

  Topics: 118

  Replies: 586

  Been thanked: 238 times
  06/26/2019 at 9:30 AM #101423
  Nicolas

  Yes, that would be possible.

  I read 2 times the description of the concept and I think that we should be able to it.

  First we need to code the “Mark Whistler’s Wave PM” indicator. Then a loop through the 32 periods (from 10 to 600? if we refer to the screenshots?) that plot a dot for each tested period and with a colour change depending of the mass level retrieved.

  Do you have an idea of where to find the code for that indicator? (didn’t have the time to browse now..). Thanks.

  EDIT: I added a screenshot of the original indicator to your post for future reference.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/26/2019 at 12:55 PM #101447
  Nicolas

  I found out that the “Mark Whistler’s Wave PM” is like a bollinger bandwidth or kinda..

  delta = average[14] - (average[14]-std[14]*2.2) width = highest[100](delta)-lowest[100](delta) return delta /width

  1 2 3 4

  delta = average[14] - (average[14]-std[14]*2.2) width = highest[100](delta)-lowest[100](delta)   return delta /width

  So now, let’s try to make some colors! 🙂

  1 user thanked author for this post.

  Vonasi

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/26/2019 at 2:43 PM #101450
  Nicolas

  There must be some conditions to generate so much different colors in the original indicator. In this first version, I only made a simple matrix that gradient color from red (indicator result of that period is extreme)  to green.

  //delta = average[14] - (average[14]-std[14]*2.2) //width = highest[100](delta)-lowest[100](delta) // //return delta /width defparam drawonlastbaronly=false //defparam calculateonlastbars=500 dev=1.5 startperiod=10 period=startperiod while period<=600 do period=max(startperiod,period+18) delta = average[period] - (average[period]-std[period]*dev) width = highest[100](delta)-lowest[100](delta) result = min(100,(delta/width)*100) R = max(0,50+(200-(result-50)*12)) G = max(0,50+(200+(result-50)*12)) drawtext("■",barindex,period,dialog,bold,18) coloured(min(r,255),min(g,255),0) wend return startperiod,600

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

  //delta = average[14] - (average[14]-std[14]*2.2) //width = highest[100](delta)-lowest[100](delta) // //return delta /width   defparam drawonlastbaronly=false //defparam calculateonlastbars=500   dev=1.5 startperiod=10   period=startperiod while period<=600 do period=max(startperiod,period+18) delta = average[period] - (average[period]-std[period]*dev) width = highest[100](delta)-lowest[100](delta) result = min(100,(delta/width)*100) R = max(0,50+(200-(result-50)*12)) G = max(0,50+(200+(result-50)*12))   drawtext("■",barindex,period,dialog,bold,18) coloured(min(r,255),min(g,255),0) wend   return startperiod,600

  I’m not satisfied but this is just the beginning 🙂

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/26/2019 at 3:10 PM #101456
  Vonasi

  Great start Nicolas. I notice in the image of the original indicator that the y scale is not linear.  It seems that less high value periods are tested than low value periods. Once you’ve perfected the colours then perhaps the period calculation could then be tweaked to get something similar. You’d need to calculate a separate y value to replace period in the DRAWTEXT. The only problem that I see with this is that the scale to the right would be wrong so perhaps on second thoughts we’ll just have to put up with too many results!

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  06/26/2019 at 3:37 PM #101458
  Nicolas

  I think that the Y scale is logarithmic…

  But I just found that the indicator is not so similar to a BB width, because the normalization is not really dependent to the period of observation, and that’s where the originality and accuracy of the method rely, I guess I must restart from scratch! 😐

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/26/2019 at 4:01 PM #101459
  Nicolas

  This is the good one now:

  BandsPeriod=14 BandsDeviations=1.25 Chars= 100 avg= average[BandsPeriod](close) if barindex>Chars then sum=0 temp=0 for j=BandsPeriod-1 downto 0 do temp = Close[j] - avg sum = sum + temp * temp next Dev= BandsDeviations * Sqrt(sum / BandsPeriod) Dev1 = square(Dev / Pointsize) //changed temp=summation[Chars](temp+dev1) temp=sqrt(temp/Chars)*pointsize if temp<>0 then temp=dev/temp endif x=temp if(x>0) then iexp=Exp(-2*x) returnNum= (1-iexp)/(1+iexp) osc= (returnNum) else iexp=Exp(2*x) returnNum=(iexp-1)/(1+iexp) osc= (returnNum) endif endif return osc

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37

  BandsPeriod=14 BandsDeviations=1.25 Chars= 100   avg= average[BandsPeriod](close)   if barindex>Chars then   sum=0 temp=0 for j=BandsPeriod-1 downto 0 do temp = Close[j] - avg sum = sum + temp * temp next   Dev= BandsDeviations * Sqrt(sum / BandsPeriod) Dev1 = square(Dev / Pointsize)   //changed temp=summation[Chars](temp+dev1) temp=sqrt(temp/Chars)*pointsize if temp<>0 then temp=dev/temp endif   x=temp if(x>0) then iexp=Exp(-2*x) returnNum= (1-iexp)/(1+iexp) osc= (returnNum) else iexp=Exp(2*x) returnNum=(iexp-1)/(1+iexp) osc= (returnNum) endif endif   return osc

  As you can see in the attached picture, the Bollinger Bandwidth I posted previously was close (the one at the bottom), but this code is the perfect one I will use now for the heatmap.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/26/2019 at 4:51 PM #101473
  AlgoAlex

  Very interesting gentlemen.

  I’d like to help but it is far beyond my actual possibilities.

   

   

  

  AlgoAlex

  Participant

  Topics: 33

  Replies: 64

  Been thanked: 25 times
  06/26/2019 at 4:53 PM #101474
  Nicolas

  I simplified the code in order to incorporate it into the loop to make the heatmap: (trying to avoid nested loops…)

  PRC_Mark Whistler’s Wave PM

  BandsPeriod=14 BandsDeviations=1.25 Chars= 100 if barindex>Chars then Dev = std[BandsPeriod]*BandsDeviations Dev1 = square(Dev / Pointsize) temp=summation[Chars](temp+dev1) temp=sqrt(temp/Chars)*pointsize if temp<>0 then temp=dev/temp endif x=temp if(x>0) then iexp=Exp(-2*x) returnNum= (1-iexp)/(1+iexp) osc= (returnNum) else iexp=Exp(2*x) returnNum=(iexp-1)/(1+iexp) osc= (returnNum) endif endif return osc

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

  BandsPeriod=14 BandsDeviations=1.25 Chars= 100   if barindex>Chars then   Dev = std[BandsPeriod]*BandsDeviations Dev1 = square(Dev / Pointsize)   temp=summation[Chars](temp+dev1) temp=sqrt(temp/Chars)*pointsize if temp<>0 then temp=dev/temp endif   x=temp if(x>0) then iexp=Exp(-2*x) returnNum= (1-iexp)/(1+iexp) osc= (returnNum) else iexp=Exp(2*x) returnNum=(iexp-1)/(1+iexp) osc= (returnNum) endif endif   return osc

   

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/26/2019 at 6:02 PM #101475
  Vonasi

  The output is zero with any bandsperiod value above 102?

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  06/26/2019 at 6:16 PM #101476
  Nicolas

  Right! Let’s try with this fix:

  BandsPeriod=103 BandsDeviations=1.25 Chars= 100 if barindex>max(Chars,BandsPeriod) then Dev = std[BandsPeriod]*BandsDeviations Dev1 = square(Dev / Pointsize) temp=summation[Chars](temp+dev1) temp=sqrt(temp/Chars)*pointsize if temp<>0 then temp=dev/temp endif x=temp if(x>0) then iexp=Exp(-2*x) returnNum= (1-iexp)/(1+iexp) osc= (returnNum) else iexp=Exp(2*x) returnNum=(iexp-1)/(1+iexp) osc= (returnNum) endif endif return osc

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

  BandsPeriod=103 BandsDeviations=1.25 Chars= 100   if barindex>max(Chars,BandsPeriod) then   Dev = std[BandsPeriod]*BandsDeviations Dev1 = square(Dev / Pointsize)   temp=summation[Chars](temp+dev1) temp=sqrt(temp/Chars)*pointsize if temp<>0 then temp=dev/temp endif   x=temp if(x>0) then iexp=Exp(-2*x) returnNum= (1-iexp)/(1+iexp) osc= (returnNum) else iexp=Exp(2*x) returnNum=(iexp-1)/(1+iexp) osc= (returnNum) endif endif   return osc

   

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/26/2019 at 6:31 PM #101477
  Nicolas

  Another more simplified version (trying to make it works for the heatmap loop with previous version but with no succeed, so I will try now with this one).

  BandsPeriod=14 BandsDeviations=1.25 Chars= 100 if barindex>max(Chars,BandsPeriod) then Dev = std[BandsPeriod]*BandsDeviations Dev1 = square(Dev / Pointsize) temp=sqrt(average[chars](dev1))*pointsize if temp<>0 then temp=dev/temp endif if(temp>0) then iexp=Exp(-2*temp) returnNum= (1-iexp)/(1+iexp) osc= (returnNum) else iexp=Exp(2*temp) returnNum=(iexp-1)/(1+iexp) osc= (returnNum) endif endif return osc

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

  BandsPeriod=14 BandsDeviations=1.25 Chars= 100   if barindex>max(Chars,BandsPeriod) then   Dev = std[BandsPeriod]*BandsDeviations Dev1 = square(Dev / Pointsize)   temp=sqrt(average[chars](dev1))*pointsize if temp<>0 then temp=dev/temp endif   if(temp>0) then iexp=Exp(-2*temp) returnNum= (1-iexp)/(1+iexp) osc= (returnNum) else iexp=Exp(2*temp) returnNum=(iexp-1)/(1+iexp) osc= (returnNum) endif endif   return osc

   

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/26/2019 at 6:46 PM #101478
  jebus89

  Amazing work Nicolas, thank you so much for ur time so far.

  

  jebus89

  Participant

  Topics: 118

  Replies: 586

  Been thanked: 238 times
  06/26/2019 at 7:16 PM #101480
  Nicolas

  Better now, but still not good… result of each iteration of the indicator in the loop does not match the exact same value as the standalone indicator (!?).

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  06/27/2019 at 1:50 PM #101534
  Nicolas

  Still in debugging mode trying to find out why values of the indicator are not correctly calculated in the loop.

  I found a complete explanation about the color matrix used for the 3D heatmap:

  This is an extension of the WAVE-PM indicator as explained by Mark Whistler in his book Volatility Illuminated. 
  This heatmap represent different lengths of WAVE-PM indicator, starting from 20 and incremented by 15 until 485. 
  As a reminder, WAVE-PM compare the size of the current distribution to the last 100. 
  The result is represented as a number between 0 and 1. 
  The more close to 0, the more contracted the distribution is. 
  The more close to 1, the more expanded the distribution is. 

  The color code is : 
  Between 0.35 and 0.5 also known as the “Gear change” level => Grey 
  Between 0.5 and 0.7 also known as the “Consolidation” level => Green 
  Between 0.7 and 0.9 also known as the “Breakout” level => Blue 
  Between 0.9 and 1 also known as the “Danger” level => Red 
  Note that i have colored results below 0.35 as black 
  Note also that each level has been divided in 2 colors : The light one for the lower half of the range and the dark one for the higher half. 

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
• Author
  Posts