The Noise Elimination Technology Indicator was created by John Ehlers (Stocks and Commodities Dec 2020 pg 17) and he created this indicator to be used with his version of RSI but I think it works well with any price data or any indicator really.
I'm trying a new signal system due to a request from @luckyCamel58789 so let me know what you think. I now differentiate between a buy and a strong buy when the indicator increases over itself twice and vice versa. Dark green is a strong buy and light green is a regular buy. Dark red is a strong sell and light red is a regular sell.
Let me know what indicators you would like to see me publish!
I've been playing around with this and I noticed changing inp to src and adding resolution to the study properties removes a 1 bar lag, but as that's the same as the original with repainting. I didn't know the standard behavior causes repainting and makes me wonder what other indicators I have which may be unintentionally repainting causing unrealistic backtesting?
Also a useful tweak I like to make to indicators like this and fisher transforms is using a variable rather than fixed 1 bar offset for the trigger. It adds more flexibility and often reduces false signals as shown in the image below.
i.e.
slo = net - nz(net(offset)) *square brackets don't show in TW comments.
A big reason I like to add optional net smoothing to other indicators is it can turn them into oscillators simplifying buy/sell signals.
cheatcountry
⋅
@AtomOfScent, yeah around 90% of the indicators on here repaint by default which I'm surprised more users don't notice. yeah I'm still experimenting with this new signal system but I can add a variable to allow changes
AtomOfScent
⋅
@cheatcountry, wow. If this is the case the default behaviour should not be repainting. As ~90% of indicators won't be recoded to fix it and barely anyone creating them even knows about this, @PineCoders should add a standard setting for all indicators.
arankioy
⋅
fascinating my friend, works on 1 minute chart as a gem
I converted this to a function a long time ago making it easy to integrate in other scripts, may as well share here (not sure if the result is the same as it sometimes leads to unexpected results):
f_net(src, len)=>
denom = len * (len - 1) / 2
num = 0.0
for i = 1 to len - 1
for k = 0 to i - 1
num := num - sign(src - src)
num/denom
cheatcountry
⋅
@AtomOfScent, A couple of reasons why you are getting unexpected results. Ehlers wrote it to be used with an array. Also i is supposed to be for i = 2 to len and k is supposed to be k = 1 to i - 1
AtomOfScent
⋅
@cheatcountry, thank you, I'll make those tweaks, and also converting your version to a function. :)