Regression Channel (ShareScope-style, parallel)

What it does

Replicates ShareScope’s Trend of displayed data look: a single straight linear-regression line (dashed) across a chosen window with parallel, constant-width bands above and below, plus optional shading.

Use it to see the overall trend gradient for a period and a statistically sized channel based on the fit’s residual error.

How it works (math, short)

Computes an OLS regression once over the analysis window.

Residual standard error s is derived from SSE and degrees of freedom (n−2).

Band half-width is constant across the window:

Mean CI (narrower): half = z * s / √n

Prediction (wider): half = z * s * √(1 + 1/n)

Three straight, parallel lines are drawn from the regression endpoints; midline is dashed.

This is intentionally not a tapered CI (which widens at the ends). It matches the visual behaviour of ShareScope’s shaded trend line channel.

Inputs

Source – Price series (Close, High, Low, HL2, etc.).

Use last N bars / N (bars) – Rolling window length.

From / To (date mode) – Alternative fixed date window.

Confidence (%) – 90 / 95 / 99 / Custom (uses z≈t).

Custom Z (t) – Override the quantile if desired.

Prediction bands – Use wider prediction envelope instead of mean CI.

Shade region + colors / opacity / line width.

Usage

To mimic ShareScope exactly, pick the same date span (use date mode) and set Confidence 99%.

Choose Prediction OFF for a tighter “confidence” look; ON for a wider, more permissive channel.

If ShareScope used High as source, set Source = High here as well.

Notes & limitations

TradingView does not expose the visible viewport to Pine. The script cannot auto-read “displayed data.” Use last N bars or date range.

Bands are parallel by design. Prices may close outside; the channel does not bend.

Window capped at 5,000 bars for performance. No alerts are emitted.

Differences vs TV’s native tools

Linear Regression (drawing) – manual object; no statistical sizing or shading.

Linear Regression Channel (indicator) – uses price standard deviations around the regression; width is a user stdev multiple.

This script – uses residual error of the OLS fit and a z/t quantile to size a statistically meaningful parallel channel.

Changelog

r3.1 – Guard fix (no return at top level), minor refactor, stable line updates.

r3 – Switched to single-fit OLS with parallel constant-width bands (ShareScope look).

(Earlier experimental builds r1–r2.2 implemented rolling/tapered CI; superseded.)

Disclaimer: Educational use only. Not investment advice.