Schiff variants change the origin of a pitchfork’s median line. That changes geometry, not the probability that price will respect it. We use the names in TradingView’s Auto Pitchfork documentation: Schiff changes the price coordinate; Modified Schiff changes both price and bar coordinates. Calculate the line before deciding what a test of it should measure.

What the Schiff Pitchfork is

For three selected points A, B and C, the classic median joins A to the midpoint M of B–C. Our model’s two rails are parallel to that median and pass through B and C. They are not standard deviations, confidence bounds or observed buy/sell orders.

For Schiff, S keeps A’s bar index and takes the mean of A and B’s prices. For Modified Schiff, MS is the midpoint of A–B in both coordinates. Both medians pass through M. Price coordinates alone do not determine slope; the bar coordinates and price scale must be recorded.

Distance from a line needs a specified metric

The same observation can be described by a vertical difference or a perpendicular distance. Those numbers represent different geometric constructions.

Vertical and perpendicular distance from a lineOn arbitrary axes, y = 0.75x + 100 gives 109 at x = 12. P = (12; 114) is 5 units vertically above it. Perpendicular projection Q = (14.4; 110.8) gives distance 4. Vertical projection V = (12; 109) is not the nearest point.VERTICAL 5, PERPENDICULAR 401020100110120P (12; 114)V (12; 109)Qy = 0.75x + 100Arbitrary axes: 1 unit = 20 px on both.At x = 12: 114 − 109 = 5.Perpendicular: 5/√1.5625 = 4.

Fictional model on normalized arbitrary axes. The line y = 0.75x + 100 is given; we do not choose pitchfork pivots here. P = (12; 114). One unit on either drawn axis spans 20 px. Price and time have no natural shared Euclidean distance; changing normalization changes this metric. The document TradingView — Schiff pitchfork drawing tool confirms that tool coordinates include price and bar number; the school distance on normalized axes is our addition.

Both arbitrary axes use equal scale to show school geometry; this does not give price and time a natural common market unit.

The vertical difference compares point and line at the same x: 114 − 109 = 5. The shortest geometric segment ends at Q = (14.4; 110.8), with components 2.4 and −3.2 and length 4. If a rule says “near the median,” specify price difference at a given time or a chosen normalized metric. Correct arithmetic does not establish pitchfork performance or an advantage from a new filter.

Standard and Modified Schiff — where the difference lies

Schiff comes in two flavours that differ in how many axes you shift the starting point in. In the Standard Schiff the median origin moves vertically only — in price: it lands halfway between the prices of P0 and P1, but stays on the same candle as P0. In the Modified Schiff you move it vertically and horizontally at once, so the origin lands exactly at the midpoint of the line joining P0 and P1, in both price and time.

Documentation labels are not entirely consistent: TradingView’s older plain-Schiff drawing page describes a two-axis shift, while Auto Pitchfork distinguishes it from Modified Schiff. We follow the latter definitions here. Record the actual coordinates, not merely the menu label. Selecting whichever line fits later turns best would use information the test was supposed to evaluate.

When to use Schiff and when the classic Andrews fork

Use a hypothetical linear-price model with equally spaced bar indices: A = (10;100), B = (14;112), C = (18;104). Then M = (16;108), S = (10;106) and MS = (12;106). Classic slope is (108−100)/(16−10) = 4/3; Schiff is (108−106)/(16−10) = 1/3; Modified is (108−106)/(16−12) = 1/2 price units per bar. At bar 22 the medians are 116, 110 and 111 respectively. Modified is steeper than Schiff here, not gentler.

The choice must be fixed before observing later prices. The articles on trend-following systems and the Andrews fork give related context; none of these lines becomes a profitable signal just because its drawn angle looks more comfortable.

How to draw and trade the Schiff Pitchfork

For a selected variant with slope m, the B rail is 112 + m(t−14) and the C rail is 104 + m(t−18). At t = 22, the classic lower/median/upper values are approximately 109.3333 / 116 / 122.6667; Schiff gives 105.3333 / 110 / 114.6667; Modified gives 106 / 111 / 116. In each row the median is the mean of the two rails. Schiff’s vertical width is 28/3, about 9.3333 price units, not a stop distance or expected volatility.

Holding A and C prices and all times fixed, changing B’s price shifts each Schiff or Modified median vertically without changing its slope. Algebra gives mSchiff = (Cprice−Aprice)/(tB+tC−2tA) and mModified = (Cprice−Aprice)/(tC−tA). For Bprice 110,112,114, the Schiff/Modified medians at t22 are 109/110, 110/111 and 111/112. With tA < tB < tC and unequal A/C prices, Modified has greater absolute slope than Schiff under these linear-coordinate definitions. Equal A/C prices make both horizontal. This is geometry, not proof of support or liquidity: see draw support and resistance.

For a separate EUR/USD geometry exercise, keep P0 = (0;1.0750), P1 = (4;1.0900), P2 = (8;1.0820). M = (6;1.0860). The two-axis origin MS = (2;1.0825) gives Modified slope 0.0035/4 = 0.000875 per bar. The one-axis Schiff origin S = (0;1.0825) instead gives slope 0.0035/6. This is a second hypothetical point set, not a quoted market or trade result.

An honest caveat — this is interpretation, not a certainty

Suppose each chosen price pivot must be strictly above or below the relevant extremes of two bars on either side; equality fails. Given that A, B and C meet this rule, they become known after bars 12, 16 and 20 close. The complete fork is therefore available after 20, not at C’s bar 18. A plot extended backwards does not alter availability. Store each old construction when a new triplet is confirmed.

Freeze the point-selection rule, variant, coordinate scale and observation start before testing. Define “near a line”: a vertical price difference at the current bar is different from a normalized perpendicular distance, as the diagram shows. Stretching screen axes changes the drawn angle. The geometry supplies no execution price, entry or exit by itself. The technical analysis section offers related methods; compare a complete rule after costs rather than assuming extra filters produce an edge.

What to do tomorrow

  1. Calculate M, S and MS for the same recorded triplet, including bar indices and a linear price scale. Check that each median passes through M.
  2. For the first model, change only Bprice by 2 units. Check the unchanged slopes and one-unit median shift; do not switch variants after seeing later reactions.
  3. For Schiff m = 1/3, calculate the C rail at bar 24: 104 + (24−18)/3 = 106. That is a line level, not a forecast of a bounce.
  4. Record extreme time and confirmation time separately. Specify a full entry/exit and monetary-risk rule if testing trades; a universal one-percent balance cap is not established by this drawing.