kvak wrote: Mon Nov 23, 2026 5:57 am
Hello.
I don't think you'll find anything; as I wrote, it's my experiment.
It occurred to me when I was porting two indicators from TV (where the Lorentzian is used in a different way - I post some beta versions in post below. ).
And trying apply different weighting calculation to moving average.
Based on "Machine Learning: Lorentzian Classification"
Original: © jdehorty, Mozilla Public License 2.0
v1.1+ : v1.0 core (past label, original neighbour selection)
on the user's template functions, with OPTIONAL additions
inspired by the "Quantum ML Engine" concepts:
- ATR neutral zone for training labels
- switchable volatility filter (off / fast / slow)
- switchable regime filter (off / original / trend strength)
- higher-timeframe EMA confluence filter (closed HTF bar)
- exit modes: fixed / dynamic / ATR trail / Supertrend
- MFI and Fisher Transform as optional features
- extended stats: net, profit factor, avg win/loss, max DD,
close-to-close calibration win-rate
With default settings the signals follow v1.0.