Andrews’ Pitchfork uses three selected time-and-price points to construct a median line and two parallels. It provides geometric references for describing a chart. The construction alone does not show that price will return to its centre, bounce at a tine or reach the next line.

What Andrews' Pitchfork is and how to draw it

In the standard construction, choose three anchors in time order: A, B and C, also often labelled P0, P1 and P2. An illustrative rising structure can use a low, the next high and a correction low. Each point needs a time or bar coordinate as well as a price. The geometry becomes reproducible once those coordinates and the chart scale are fixed; a swing-selection rule remains a separate decision.

Find midpoint M of segment B–C on both axes. The median is the straight line through A and M; draw parallels through B and C. On a linear price scale, M averages both the time coordinates and prices. This is neither a moving average of all closing prices nor a regression fitted to the whole series. MetaTrader 5 documents the three Date/Value anchors and distinguishes moving the handle’s beginning from shifting the whole object. TradingView exposes prices and bar numbers and a choice of pitchfork variant. Check the actual platform controls rather than assume one click sequence works everywhere.

The median-return hypothesis and what geometry establishes

Interpreting a tine as support or resistance, or the median as a future target, is a hypothesis to test. Geometry can tell you where a line lies at a given time; it cannot establish the frequency of bounces, a healthy trend or the probability of a return. Preserve cases where price ignores the lines as well as examples that visually fit them.

The same distinction applies to horizontal support and resistance levels. A broken line need not become resistance or support automatically. Define what counts as a touch, rejection or break, and use data that were available when the setup was identified. The drawing’s interpretation is not an independently established market law.

The pitchfork's slope depends on both the prices and timestamps of its three anchors. Parallel boundaries organise the chart; they do not guarantee a return to the median line.

How to trade reactions at the pitchfork lines

A test can examine reactions at the lower tine, median or upper tine as separate rules. A touch does not itself specify entry side, fill time, position size or exit. Comparing rule variants requires consistent costs and data rather than claiming that one line usually invites adding to a position and another usually invites profit-taking.

If a rule waits for a reversal candle, record when that candle closes and when an order could actually execute. Specify the invalidation condition and stop price separately. “Beyond the lower tine” is incomplete because the line slopes: state the time at which its price is calculated and whether/how an order is later amended. Chosen targets on other lines need their own timing and fill rules. An ordinary stop can fill worse, and the next tine is not a guaranteed destination.

Some implementations allow extra parallel reference lines. MetaTrader 5 documents signal-line levels; TradingView allows additional lines and variant settings. Record the selected levels instead of assuming one equally spaced default in every platform. Additional lines do not turn an original-channel break into evidence that price will stop at the next reference.

A hypothetical example step by step

Use a fictional EUR/USD chart with a linear price scale and equally spaced bars numbered t. Set A = (0, 1.1000), B = (4, 1.1120) and C = (8, 1.1060). Midpoint M = (6, 1.1090). The median slope is (1.1090 − 1.1000)/6 = 0.0015 per bar, or 15 pips per bar for a 0.0001 pip. This is a drawn line’s slope, not a predicted currency speed. The separate existing diagram uses different, deliberately collinear coordinates to show a degenerate case.

Fictional standard pitchfork; line values at selected bars
LinePrice equationt = 10t = 12
Median1.1000 + 0.0015t1.11501.1180
Through B1.1120 + 0.0015(t − 4)1.12101.1240
Through C1.1060 + 0.0015(t − 8)1.10901.1120

Each tine is 0.0060, or 60 pips, vertically from the median; total vertical width is 120 pips. These are geometric distances, not a standard deviation or predicted trading range. If C keeps price 1.1060 but moves to bar 12, M moves to (8, 1.1090), slope becomes 0.001125 and the median at bar 12 is 1.1135 rather than 1.1180. The same three prices therefore produce a different channel when their time coordinates change.

Changing only the A price is also different from moving the entire pitchfork. Keep B, C and M fixed in the previous example: p(t) = A + (1.1090 − A)t/6. All three candidate median lines pass through the same M, so they rotate around it.

Different hypothetical A prices, unchanged midpoint
A price at t = 0Median t = 3Median t = 6Median t = 12
1.09901.10401.10901.1190
1.10001.10451.10901.1180
1.10101.10501.10901.1170

Raising A by 10 pips raises the line before M but lowers it by 10 pips at t = 12. A higher anchor does not translate the whole future channel upward. These are counterfactual geometric comparisons on linear axes, not historical price paths or trades before C was known. Logarithmic scales and Schiff variants require their own construction.

When does a pitchfork have zero width?

Three chosen coordinates need not form a useful construction. This model detects a degenerate case.

Scroll sideways to see the full diagram

When does a pitchfork have zero width?Our collinear coordinates A(0,100), B(4,112), C(8,124) give midpoint M(6,118) and median slope 3. The median and parallels through B and C overlap because each has price equation 100+3t. At time 10, every price is 130 and the width is zero. Different colours and line widths occupy the same coordinates.Three anchors on one lineA (0; 100) • B (4; 112) • C (8; 124)100112124130ABMC046810hypothetical timeMedian: 100 + 3 tThrough B: 112 + 3(t − 4) = 100 + 3 tThrough C: 124 + 3(t − 8) = 100 + 3 t

Explicit geometric counterexample: hypothetical prices and times. A = (0; 100), B = (4; 112), C = (8; 124). These coordinates are collinear; they are not a properly selected alternating sequence of highs and lows.

The midpoint of BC is M = (6; 118), so the slope of AM is (118 − 100)/(6 − 0) = 3. The parallels through B and C have exactly the same equation as the median. At t = 10, all three indicate 130; the construction has zero width.

MT5 documentation describes the median through the midpoint of the two right anchors and the parallel lines. Computing an object from three coordinates does not establish sensible swing selection. Check the anchors before reading a “channel”.

Zero width means overlapping lines, not three independent levels.

An honest caveat: choosing the points is subjective

Two drawings can differ because their anchors, time coordinates, scale or variant differ. Point choice may involve judgement, but a fully specified construction can be reproduced. Compare the coordinates before interpreting opposite conclusions. The broader technical-analysis material separates a chosen drawing from a tested rule; avoid moving anchors only after seeing which version fits the outcome.

Compare any proposed confirmation rule separately: a Fibonacci retracement, a horizontal level or a trend-following approach does not make the combined method profitable by definition. Keep rejected cases, costs and worse-execution scenarios in the record. The Schiff pitchfork changes the construction and should be tested as a separate variant. A simulated or hindsight-selected chart is not an actual execution record.

What to do tomorrow

  1. Fix the instrument, interval, price scale, pitchfork variant and three anchor coordinates. Record a swing-selection rule and when each point becomes known.
  2. Review a later chart segment with the anchors frozen. Record line touches and breaks using definitions set before seeing the result, including cases with no apparent reaction.
  3. Change only one anchor in a separate saved drawing and calculate the line value at a stated time. Distinguish rotation around the midpoint from translation of the whole object.
  4. If testing trades, specify entry/exit quote sides, costs, risk budget and possible stop slippage. A chosen two-week demo period or one-percent limit does not establish profitability, live readiness or a guaranteed maximum loss.