cubed moving average

[MLlalala]

I am trying to write a cube weighted moving average, the logic is quite simple, instead of traditional WMA whic is 4*price+3*price[1]+2*price[2]+price[3], i wanna cube the weighting, so it goes like 64*price+27*price[1]+8*price[2]+price[3]

Here is my code, but i am at a loss how to do it correctly

input price = hl2;
input length=10;

def weight=power(length,3);
plot CWMA = Sum(weight * price, Length) / Sum(weight, Length);

Much appreaciated if anyone could help out!

 

[halcyonguy]

I am trying to write a cube weighted moving average, the logic is quite simple, instead of traditional WMA whic is 4*price+3*price[1]+2*price[2]+price[3], i wanna cube the weighting, so it goes like 64*price+27*price[1]+8*price[2]+price[3]

Here is my code, but i am at a loss how to do it correctly

input price = hl2;
input length=10;

def weight=power(length,3);
plot CWMA = Sum(weight * price, Length) / Sum(weight, Length);

Much appreaciated if anyone could help out!

you need to find the last 'length' bars on the chart.
find the last bar with a price, and count back from it, length bars. the difference of barnumber and barnumber of last bar and length, can be used to calc a series of numbers 1,2,3,...length. then use those values to calculate the weight.

i am away , so didn't verify.
i think your cwma formula will work.

def bn = BarNumber(); 
def na = Double.NaN; 

def lastbn = HighestAll( If(IsNaN(close), 0, bn )); 
def lastbar = (bn == lastbn);
def barstolast = (lastbn - bn);

input price = hl2;
input length=10;

def wt_index = length - barstolast;

def weight = if wt_index < 0 then 0 else 
power(wt_index, 3);

plot CWMA = Sum(weight * price, Length) / Sum(weight, Length);

#------------------
# test data

input test1 = no;
addchartbubble(test1, low,
wt_index + "\n" +
weight + "\n" +
price + "\n" +
CWMA
, color.yellow, no);

weighted avg
https://www.investopedia.com/terms/w/weightedaverage.asp

 

[MLlalala]

you need to find the last 'length' bars on the chart.
find the last bar with a price, and count back from it, length bars. the difference of barnumber and barnumber of last bar and length, can be used to calc a series of numbers 1,2,3,...length. then use those values to calculate the weight.

i am away , so didn't verify.
i think your cwma formula will work.

def bn = BarNumber();
def na = Double.NaN;

def lastbn = HighestAll( If(IsNaN(close), 0, bn ));
def lastbar = (bn == lastbn);
def barstolast = (lastbn - bn);

input price = hl2;
input length=10;

def wt_index = length - barstolast;

def weight = if wt_index < 0 then 0 else
power(wt_index, 3);

plot CWMA = Sum(weight * price, Length) / Sum(weight, Length);

#------------------
# test data

input test1 = no;
addchartbubble(test1, low,
wt_index + "\n" +
weight + "\n" +
price + "\n" +
CWMA
, color.yellow, no);

weighted avg
https://www.investopedia.com/terms/w/weightedaverage.asp

The code seems only plotting a limited portion. and if i just make the power function as
power(wt_index, 1); it should be equal to a normal weighted moving average. I have compared a normal weighted moving average with same length but they differ. The green line is TOS built in WMA. The parameters for both averages are price=hl2 and length =10

 

[halcyonguy]

The code seems only plotting a limited portion. and if i just make the power function as
power(wt_index, 1); it should be equal to a normal weighted moving average. I have compared a normal weighted moving average with same length but they differ. The green line is TOS built in WMA. The parameters for both averages are price=hl2 and length =10

my mistake. i was thinking that just the value on the last bar was important, so i made it to only calculates on the last 'length' bars. thinking...

 

[halcyonguy]

The code seems only plotting a limited portion. and if i just make the power function as
power(wt_index, 1); it should be equal to a normal weighted moving average. I have compared a normal weighted moving average with same length but they differ. The green line is TOS built in WMA. The parameters for both averages are price=hl2 and length =10

this will calculate a weighted average on every bar, except the first 'length' bars.
i added 2 loops, to sum the product and sum the weights. then divide them at the end.
untested.

def bn = BarNumber(); 
def na = Double.NaN; 

input price = hl2;
input length =10;
input wt_power = 3;

def wtpr_sum;
def wt_sum;
if bn <= length then {
wtpr_sum = 0;
wt_sum = 0;
} else {
wtpr_sum = fold i = 0 to length
with p
do p + ( getvalue(price, i) * power((length - i), wt_power) );

wt_sum = fold k = 0 to length
with r
do r + ( power((length - k), wt_power) );
}

plot CWMA = if wt_sum > 0 then round(wtpr_sum / wt_sum, 2) else na;
#

 

[MLlalala]

this will calculate a weighted average on every bar, except the first 'length' bars.
i added 2 loops, to sum the product and sum the weights. then divide them at the end.
untested.

def bn = BarNumber();
def na = Double.NaN;

input price = hl2;
input length =10;
input wt_power = 3;

def wtpr_sum;
def wt_sum;
if bn <= length then {
wtpr_sum = 0;
wt_sum = 0;
} else {
wtpr_sum = fold i = 0 to length
with p
do p + ( getvalue(price, i) * power((length - i), wt_power) );

wt_sum = fold k = 0 to length
with r
do r + ( power((length - k), wt_power) );
}

plot CWMA = if wt_sum > 0 then round(wtpr_sum / wt_sum, 2) else na;
#

This is perfect!!!

 

[spartan.ts]

I am trying to write a cube weighted moving average, the logic is quite simple, instead of traditional WMA whic is 4*price+3*price[1]+2*price[2]+price[3], i wanna cube the weighting, so it goes like 64*price+27*price[1]+8*price[2]+price[3]

Interesting concept. Can you explain how this MA is of value to you and where did the idea come from? thanks for your insights.

 

[MLlalala]

Interesting concept. Can you explain how this MA is of value to you and where did the idea come from? thanks for your insights.

Yes i would be very happy to discuss this idea. For traditional WMA, the weight is assigned linearly like 4321, then divided by the sum, in a 4 bar WMA, current data is "diluted" by previous data and this linear relationship is not really true. So i just use a quick way to change the relationship by giving more weight to most recent data and less for previous data. You can play with square or square root too. Hope this helps