zeonta.trend_channel() — Linear-regression trend line with standard-deviation channel bands.
What it measures
“Is this an uptrend?” is usually answered by eye. A least-squares fit answers it with a number: the slope. The channel bands around that fit show how tightly price has been hugging the trend.
Formula
Linear regression over length n bars (x = 0..n-1, y = close): slope b = (nSxy - SxSy) / (nSx^2 - (Sx)^2); intercept a = (Sy - bSx) / n; regression line = a + b*x. Channel bands = regression line +/- (multiplier x standard deviation of closes from the regression line).
Parameters
Required inputs: close
| Parameter | Default |
|---|---|
length |
100 |
multiplier |
2.0 |
Returns
| Column |
|---|
LRCM_100 |
LRCU_100 |
LRCL_100 |
LRCSLOPE_100 |
Usage
Examples run against the 300-bar OHLCV fixture in tests/data/ohlcv.csv, loaded as df. The output shown is the real output.
import pandas as pd
import zeonta
df = pd.read_csv('tests/data/ohlcv.csv', parse_dates=['date']).set_index('date')
zeonta.trend_channel(df['close'], length=50).tail(3)
LRCM_50 LRCU_50 LRCL_50 LRCSLOPE_50
date
2024-10-25 90.207156 91.314703 89.099609 -0.054643
2024-10-26 90.072080 91.214641 88.929520 -0.057086
2024-10-27 89.891669 91.106883 88.676454 -0.060957
Accessor form: df.zta.trend_channel(...)
How to read it
LRCSLOPE is the per-bar drift: positive is an uptrend, negative a downtrend, and its magnitude is the trend’s steepness. Price near LRCU is extended relative to the trend; near LRCL it is lagging behind it. The band width is the scatter of price about the fitted line, not about its mean, so a cleanly trending market gives a narrow channel however steep it is.
Pitfalls
The fit is recomputed every bar, so the channel repaints as new data arrives — the line you see today over past bars is not the line that existed back then. Also, a regression will happily fit a straight line through pure noise; check the slope against something like ADX before trusting it.