Andrews' Pitchfork

A three-pivot channel built from a median line and two parallels — deterministic geometry with subjective inputs.

What it is

Andrews' Pitchfork — also called the median line method — is a drawing tool attributed to Alan Hall Andrews, who taught it as part of his mid-20th-century 'Action-Reaction' course. From three chosen pivot points it projects a channel: a median line drawn from the first pivot through the midpoint of the second and third, plus two parallel lines passing through the second and third pivots. The geometry is completely deterministic — plain midpoint and slope arithmetic — which means the only judgment call, and the tool's biggest weakness, is which three pivots you pick. Two traders can draw materially different forks on the same chart and both be 'correct' by the tool's rules. The pitchfork describes the slope and width implied by three past swing points; it knows nothing about future price, and there is no verified evidence that markets are obliged to respect median lines. Treat a fork as a structured, reproducible way to draw a trend channel rather than as levels the market must honor. This is educational content only, not financial advice, and trading involves risk of loss.

How it works

Label the pivots P1 (the origin, an earlier swing high or low), and P2 and P3 (the next two alternating swings). Compute the midpoint M of the P2–P3 segment: M = ((x2 + x3) / 2, (y2 + y3) / 2), using bar index for x and price for y. The median line is the ray from P1 through M, with slope (My − P1y) / (Mx − P1x) in price per bar. The upper and lower parallels pass through P2 and P3 with exactly the same slope, so the fork is a channel whose width is fixed by the vertical distance between P2 and P3, extended to the right indefinitely. That is the entire construction — midpoint math, one slope, three lines. Variants change only the anchor: the Schiff pitchfork shifts P1 halfway toward P2 in price, and the modified Schiff shifts it halfway in both time and price; the parallel construction is identical. One timing detail matters for anyone coding it: the fork can only be drawn after P3 is a confirmed swing point, which typically means several bars after the actual extreme. An auto-drawing script that anchors to unconfirmed pivots will repaint as those pivots move.

How traders read it

• The median line is the center of the drawn channel. Andrews taught that price tends to return toward it — read that as a description of how channels are constructed, not a law of markets.
• The upper and lower parallels are the channel edges implied by P2 and P3; price trading near an edge describes position within the drawn channel, nothing more.
• The median line's slope is a compact summary of the trend defined by your three chosen pivots — different pivots, different trend.
• A fork has no memory: when new swings form, redrawing produces an entirely different channel, so any reading is conditional on the pivot choice staying fixed.
• Only use confirmed pivots. A swing point is not known until enough bars have printed on both sides of it, so a live fork is always drawn with a delay.
• Be skeptical of textbook examples: forks shown in educational material were selected in hindsight precisely because price appeared to respect them.

Common settings

The classical tool has no numeric parameters — the 'settings' are the three pivots themselves. Auto-drawing implementations expose a pivot length (bars required on each side of a swing to confirm it, commonly 5-15) and a variant selector: standard, Schiff, or modified Schiff. Longer pivot lengths give more stable anchors but confirm later; shorter lengths redraw more often. State your pivot rules explicitly if you want two people (or two scripts) to draw the same fork.

Strengths

• Fully deterministic geometry once the pivots are fixed — the same three points always produce the same fork, which makes it reproducible and codeable.
• Encodes both trend slope and channel width in a single construction instead of two freehand lines.
• Forces an explicit trend definition: you must say which swings define the move you are analyzing.
• The Schiff and modified Schiff variants provide systematic alternatives rather than arbitrary manual adjustments.

Pitfalls to watch

• Pivot subjectivity dominates: the math is exact but the inputs are a choice, so the tool's apparent precision can disguise an arbitrary starting decision.
• Hindsight bias: forks that 'worked' are the ones that get shown and remembered; nothing about the construction makes future respect of the lines likely.
• Late availability: P3 must be a confirmed swing, so a valid fork often cannot be drawn until well after the move it describes has begun.
• Repaint risk in auto-drawing versions: scripts that anchor to unconfirmed pivots silently move the whole fork as new bars form.
• Self-referential levels: there is no verified mechanism by which a median line attracts price; any apparent respect may be coincidence or self-fulfilling attention.
• Redrawing temptation: when price ignores a fork it is easy to pick new pivots until one fits — which turns the tool into curve-fitting.

Pine v6 example

//@version=6
indicator("Andrews Pitchfork Example", overlay = true, max_lines_count = 3)

aBack = input.int(60, "Pivot P1 bars back (swing low)", minval = 3)
bBack = input.int(40, "Pivot P2 bars back (swing high)", minval = 2)
cBack = input.int(20, "Pivot P3 bars back (swing low)", minval = 1)

var line medianLn = na
var line upperLn = na
var line lowerLn = na

if barstate.islast and bar_index > aBack
    ax = bar_index - aBack
    bx = bar_index - bBack
    cx = bar_index - cBack
    ay = low[aBack]
    by = high[bBack]
    cy = low[cBack]
    midX = (bx + cx) / 2
    midY = (by + cy) / 2
    line.delete(medianLn)
    line.delete(upperLn)
    line.delete(lowerLn)
    medianLn := line.new(ax, ay, midX, midY, extend = extend.right, color = color.teal, width = 2)
    upperLn := line.new(bx, by, bx + (midX - ax), by + (midY - ay), extend = extend.right, color = color.gray)
    lowerLn := line.new(cx, cy, cx + (midX - ax), cy + (midY - ay), extend = extend.right, color = color.gray)

Pro tip: Write down your pivot-selection rule before drawing — for example, 'confirmed swings with N bars on each side, most recent three alternating swings' — and stick to it. A fork drawn from a fixed rule is at least testable; a fork adjusted until it fits recent price is a story told backwards. Remember that the geometry being exact does not make the levels meaningful. Educational information only, not financial advice; no drawing tool predicts price, and trading involves risk of loss.

Educational only — not financial advice, not a recommendation to trade. No indicator is predictive; trading involves substantial risk of loss.