MMI Indicator Translation

• 07/02/2017 at 10:55 AM #39572
  Pietro Fontana

  Hello people,

  can someone convert this indicator in ProRealCode? I’m having difficulties in translating the median price calculation, but maybe i’m missing something easy here.

  Link of the article and how to use this indicator here: http://www.financial-hacker.com/the-market-meanness-index/

   

  var MMI(var *Data,int Length)
  {
  var m = Median(Data,Length);
  int i, nh=0, nl=0;
  for(i=1; i<Length; i++) {
  if(Data[i] > m && Data[i] > Data[i-1])
  nl++;
  else if(Data[i] < m && Data[i] < Data[i-1])
  nh++;
  }
  return 100.*(nl+nh)/(Length-1);
  }

   

  Thanks.

  Best regards.

  

  Pietro Fontana

  Participant

  Topics: 5

  Replies: 19

  Been thanked: 0 times
  07/03/2017 at 9:26 AM #39647
  Nicolas

  This code is from Zorro platform. Do you have documentation of how the Median() function is calculated? It should help a lot 🙂

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4773 times
  07/13/2017 at 12:39 PM #40564
  gabri

  Nicolas,

  I found this one in tradingview:

  //@version=2
  study(“MMI-Price Changes”)
  // Set Inputs
  inputLength=input(300)
  topband=input(76)
  bottomband=input(74)

  // Get Average Price Change
  mean(price1, price2, length) =>
  sum = 0
  for i = 0 to length
  if price1[i] – price2[i] > 0
  sum := sum + (price1[i] – price2[i])
  else
  sum := sum + (price2[i] – price1[i])
  sum / length

  // Market Meaness Index Function
  MMI(price1,price2, length) =>
  m = mean(open,close,inputLength)
  nh = 0
  nl = 0
  for i = 1 to length
  if price1[i] – price2[i] > 0
  if (price1[i] – price2[i]) > m
  if (price1[i] – price2[i]) > (price1[i+1] – price2[i+1])
  nl := nl + 1
  else
  if (price2[i] – price1[i]) < m
  if (price2[i] – price1[i]) < (price2[i+1] – price1[i+1])
  nh := nh + 1
  100 – 100*(nl + nh)/(length – 1)

  u=plot(topband)
  l=plot(bottomband)
  t=plot(90)
  b=plot(60)
  fill(u,l,black)
  fill(t,u,yellow)
  fill(b,l,blue)
  plot(MMI(open,close,inputLength))

  

  gabri

  Participant

  Topics: 6

  Replies: 116

  Been thanked: 39 times
  07/14/2017 at 11:44 AM #40630
  gabri

  Pietro,

  try to see if this code produces the same results of the code you wanted:

  length=300 rangeday=open-close a=abs(rangeday) m=summation[length](a)/length if rangeday>0 and rangeday>rangeday[1] and rangeday>m then b=1 else b=0 endif if rangeday>0 and rangeday<rangeday[1] and rangeday<m then b=1 endif c=summation[length](b) ret=100 - 100*(c)/(length - 1) return ret

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

  length=300   rangeday=open-close a=abs(rangeday) m=summation[length](a)/length   if rangeday>0 and rangeday>rangeday[1] and rangeday>m then b=1 else b=0 endif if rangeday>0 and rangeday<rangeday[1] and rangeday<m then b=1 endif   c=summation[length](b)   ret=100 - 100*(c)/(length - 1) return ret

   

  

  gabri

  Participant

  Topics: 6

  Replies: 116

  Been thanked: 39 times
  07/14/2017 at 8:43 PM #40672
  Pietro Fontana

  Hello guys,

  sorry for late response, but i forgot to subscribe the thread 🙂

  Yes Nicolas, the code is from Zorro, but i did not find anything more in their forum sadly.

  Thanks Gabri for the code, i will try it asap.

  

  Pietro Fontana

  Participant

  Topics: 5

  Replies: 19

  Been thanked: 0 times
  07/14/2017 at 10:49 PM #40682
  Pietro Fontana

  Hi Gabri,

  i’ve tried the code, there’s something that is not working.

  The indicator value should be between 0 and 100, because the idea is that if the value > 75 there’s no trend, below 75 is trending.

  See the attached image, DAX daily, we have negative value for the indicator, -8000 etc.

  It seems that the second part of the formula ret=100 – 100*(c)/(length – 1) is producing a value far high than 100, it should be <= 100.

  In particulare, variable C have value of 25000 or more on this instruments. So the problem should rely on the variable C in my opinion, but i don’t found why.

  

  Pietro Fontana

  Participant

  Topics: 5

  Replies: 19

  Been thanked: 0 times
  07/15/2017 at 12:59 PM #40720
  gabri

  Pietro,

  this is what I get with a daily DAX (10000 units):

   

  

  gabri

  Participant

  Topics: 6

  Replies: 116

  Been thanked: 39 times
  07/15/2017 at 1:04 PM #40722
  gabri

  The highest value I got for c is 175 (once) other than that it never went above 100. I saw you called the code [beta]MMI, maybe you changed some line inadvertently.

  

  gabri

  Participant

  Topics: 6

  Replies: 116

  Been thanked: 39 times
  07/16/2017 at 12:55 PM #40765
  Pietro Fontana

  Hi Gabri,

  you are right, your code is ok, but i’ve only added a colouring logic and put the length in a variable, the code is this:

   

  //length=300 rangeday=open-close a=abs(rangeday) m=summation[length](a)/length if rangeday>0 and rangeday>rangeday[1] and rangeday>m then b=1 else b=0 endif if rangeday>0 and rangeday<rangeday[1] and rangeday<m then b=1 endif c=summation[length](b) MMI = 100 - 100*(c)/(length - 1) IF MMI > MMI[1] THEN R = 0 G = 255 B = 0 ENDIF IF MMI < MMI[1] THEN R = 255 G = 0 B = 0 ENDIF IF MMI = MMI[1] THEN R = 255 G = 255 B = 255 ENDIF RETURN MMI COLOURED(R,G,B) AS "MMI"

  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 38

  //length=300   rangeday=open-close a=abs(rangeday) m=summation[length](a)/length   if rangeday>0 and rangeday>rangeday[1] and rangeday>m then b=1 else b=0 endif if rangeday>0 and rangeday<rangeday[1] and rangeday<m then b=1 endif   c=summation[length](b)   MMI = 100 - 100*(c)/(length - 1)   IF MMI > MMI[1] THEN R = 0 G = 255 B = 0 ENDIF   IF MMI < MMI[1] THEN R = 255 G = 0 B = 0 ENDIF   IF MMI = MMI[1] THEN R = 255 G = 255 B = 255 ENDIF   RETURN MMI COLOURED(R,G,B) AS "MMI"

  By the way, since i want to use the MMI in automated strategy, i will leave it plain for now and see if we got some performance improvement in trend following strategy.

  Many thanks.

  Regards.

  

  Pietro Fontana

  Participant

  Topics: 5

  Replies: 19

  Been thanked: 0 times
  07/16/2017 at 1:48 PM #40772
  gabri

  Prova a modificare le seguenti righe cosi:

  IF MMI > MMI[1] THEN Red = 0 Green = 255 Blue = 0 ENDIF IF MMI < MMI[1] THEN Red = 255 Green = 0 Blue = 0 ENDIF IF MMI = MMI[1] THEN Red = 255 Green = 255 Blue = 255 ENDIF RETURN MMI COLOURED(Red,Green,Blue) AS "MMI"

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

  IF MMI > MMI[1] THEN Red = 0 Green = 255 Blue = 0 ENDIF IF MMI < MMI[1] THEN Red = 255 Green = 0 Blue = 0 ENDIF IF MMI = MMI[1] THEN Red = 255 Green = 255 Blue = 255 ENDIF RETURN MMI COLOURED(Red,Green,Blue) AS "MMI"

   

   

  1 user thanked author for this post.

  Nicolas

  

  gabri

  Participant

  Topics: 6

  Replies: 116

  Been thanked: 39 times
  07/16/2017 at 2:28 PM #40780
  Pietro Fontana

  Funziona, mi sono accorto di aver utilizzato la variabile B per due cose differenti, credo fosse quello l’errore giusto?

  

  Pietro Fontana

  Participant

  Topics: 5

  Replies: 19

  Been thanked: 0 times
  07/16/2017 at 2:30 PM #40781
  gabri

  Si, capita spesso non preoccuparti 🙂

  

  gabri

  Participant

  Topics: 6

  Replies: 116

  Been thanked: 39 times
  07/16/2017 at 6:51 PM #40808
  Nicolas

  Does it working as expected  please Pietro?

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4773 times
  07/16/2017 at 11:33 PM #40838
  Pietro Fontana

  Hi Nicolas,

  to be honest i still can’t say if this is working as expected, since i can’t test the original code from jcl that is in a C variant.

  Said that, the code from Gabri seems to produce a reasonable result. I’ve read that jcl smooth the MMI indicator before use it in strategy, look here: http://www.financial-hacker.com/boosting-systems-by-trade-filtering/

  So i’ve smoothed it with Supersmoother (taken from the thread you make), but maybe a low pass filter can be used, it seems that jcl used a low pass filter, but since he recall a function i can’t say what’s inside.

  This is the code for the smoother MMI that i’ve made:

  length=300 rangeday=open-close a=abs(rangeday) m=summation[length](a)/length if rangeday>0 and rangeday>rangeday[1] and rangeday>m then b=1 else b=0 endif if rangeday>0 and rangeday<rangeday[1] and rangeday<m then b=1 endif c=summation[length](b) ret=100 - 100*(c)/(length - 1) Period = 10 Data = ret PI = 3.14159 f = (1.414*PI) / Period a = exp(-f) c2 = 2*a*cos(f) c3 = -a*a c1 = 1 - c2 - c3 if barindex>length then S = c1*(Data[0]+Data[1])*0.5 + c2*S[1] + c3*S[2] endif return S

  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

  length=300   rangeday=open-close a=abs(rangeday) m=summation[length](a)/length   if rangeday>0 and rangeday>rangeday[1] and rangeday>m then b=1 else b=0 endif if rangeday>0 and rangeday<rangeday[1] and rangeday<m then b=1 endif   c=summation[length](b)   ret=100 - 100*(c)/(length - 1)     Period = 10 Data = ret PI = 3.14159 f = (1.414*PI) / Period a = exp(-f) c2 = 2*a*cos(f) c3 = -a*a c1 = 1 - c2 - c3 if barindex>length then S = c1*(Data[0]+Data[1])*0.5 + c2*S[1] + c3*S[2] endif return S

  Another things that i can’t reproduce is the Peak and Valley function, it should be something related to high and low of the period.

  In the linked article jcl seems to not consider the value higher or lower 75, but if MMI is lowering or raising. In particular he used the Falling function: DoTrade = falling(MMI_Smooth)

  So i suppose that if the MMI is lowering the price should be trending. I’ve made some raw test that did not give good result, but i don’t have a solid automated trend following system, it would be good to use this as a filter to a solid trend following system to see if it can improve it.

  I will report the result of further investigation.

  Regard.

  

  Pietro Fontana

  Participant

  Topics: 5

  Replies: 19

  Been thanked: 0 times
  07/17/2017 at 7:13 PM #40894
  gabri

  From what I saw I am not sure you can use this indicator like trigger. I believe is more like a trend defining tool (like an adx)

  

  gabri

  Participant

  Topics: 6

  Replies: 116

  Been thanked: 39 times
• Author
  Posts