Place three pivots on TradingView
Use ordinary candles, a fixed symbol and interval, and a linear price scale for the calculations below. For a rising candidate, choose a low A, a later high B and a later retracement low C. Reverse that sequence for a falling candidate. These are selected swing points, not session Pivot Points calculated from prior high, low and close.
- Choose Pitchfork from the drawing toolbar and click A, B and C in chronological order. In its Coordinates settings, verify the bar and price of each point instead of relying on approximate mouse placement.
- Start with the median and the two parallels through B and C. Extra ratio lines are optional and must be recorded if they form part of a rule.
- Select the desired variant in Style and save the anchor coordinates, chart interval and scale with the screenshot. Changing a variant changes the geometry even when the anchors stay fixed.
- For an automatic alternative, add TradingView Auto Pitchfork and record its Depth and Type. Its documentation describes Depth 10 as using five neighboring bars on each side of a pivot. A later update can choose a new set of points; preserve the fork that existed at your decision time.
Calculate the median and both parallels
Let A = (xA, yA), B = (xB, yB) and C = (xC, yC), where x is a chart bar index and y is price. For a standard fork, M = ((xB + xC) / 2, (yB + yC) / 2). The slope is m = (yM - yA) / (xM - xA). At bar x, the median is yA + m × (x - xA), the B parallel is yB + m × (x - xB), and the C parallel is yC + m × (x - xC).
- These invented coordinates demonstrate geometry, not a market forecast. At bar 30 the core channel runs from 120 to 140, with its median at 130.
- The median is a geometric line through selected points. It is not the average closing price, a least-squares regression, or a statistical confidence interval. No return-to-median probability follows from this construction.
- Keep bar indices and elapsed clock time separate. The example advances one unit per chart bar, including across overnight or weekend gaps in trading.
| Hypothetical input or result | Calculation |
|---|
| Three confirmed anchors | A = (0, 100), B = (10, 120), C = (20, 110) |
| Midpoint M | ((10 + 20) / 2, (120 + 110) / 2) = (15, 115) |
| Standard slope | (115 - 100) / (15 - 0) = 1 price point per bar |
| Median at bar 30 | 100 + 1 × (30 - 0) = 130 |
| B parallel at bar 30 | 120 + 1 × (30 - 10) = 140 |
| C parallel at bar 30 | 110 + 1 × (30 - 20) = 120 |
Compare standard, Schiff and Modified Schiff
TradingView Auto Pitchfork documents Schiff as shifting the median origin halfway from A toward B in price only. Modified Schiff shifts it halfway in both price and time. The table uses those explicit definitions and keeps M, B and C unchanged. Recalculate the slope from the chosen origin to M, then draw the parallels through B and C with that slope.
- Schiff is flatter in this example; Modified Schiff is between Schiff and standard. Do not assume one variant is always the flattest for every set of anchors.
- Check the actual median origin and line values when comparing tools. A variant name alone is insufficient if implementations use different coordinate conventions.
- Choose the variant before evaluating future touches. Switching after every miss makes the historical fit look better without testing a fixed rule.
| Variant | Median origin | Slope | Levels at bar 30: lower / median / upper |
|---|
| Standard Andrews | A = (0, 100) | 15 / 15 = 1 | 120 / 130 / 140 |
| Schiff | (xA, (yA + yB) / 2) = (0, 110) | 5 / 15 = 1/3 | 113.3333 / 120 / 126.6667 |
| Modified Schiff | ((xA + xB) / 2, (yA + yB) / 2) = (5, 110) | 5 / 10 = 0.5 | 115 / 122.5 / 130 |
A confirmed pivot arrives after its plotted location
For a reproducible example, define a pivot low as strictly lower than the lows of two bars on each side; define highs symmetrically. Reject ties and missing comparison bars. Assume A and B have already qualified. If C is at bar 20, its right-side comparison requires bars 21 and 22 to close. The completed fork becomes available at the close of bar 22, not at bar 20.
- Record pivot bar 20 and confirmation bar 22 separately. A historical label drawn back on bar 20 does not make the fork available there in real time.
- For this exercise, evaluate breaks starting with bar 23. If bar 22 instead makes a low of 109, the candidate at bar 20 fails and this fork never becomes active.
- This strict two-bar rule is a stated research choice, not a claim about every automatic indicator. Verify tie handling, pivot replacement and confirmation timing in the actual implementation.
| Bar | Hypothetical low | Information available |
|---|
| 18 and 19 | 113 and 112 | The left side of C is known. |
| 20 | 110 | Candidate low only; later lows are unknown. |
| 21 | 111 | One right-side bar has closed. |
| 22 | 112 | Both right-side bars have closed above 110; C qualifies. |
Distinguish a touch, a close outside and invalidation
Freeze the standard example once C is confirmed. At bar 23, its lower parallel is 110 + (23 - 20) = 113 and its upper parallel is 120 + (23 - 10) = 133. Define a lower break as a completed close strictly below the lower parallel minus a chosen nonnegative price buffer. An upper break uses a close strictly above the upper parallel plus that buffer.
- A strict event flag describes an observation, not a reversal prediction. A later close can return inside the channel.
- One possible study policy retires the fork after its first qualifying close outside. Record that event before selecting new anchors. Another policy may require two closes, but that must be specified beforehand and each close compared with its own bar-specific line value.
- Store the first break timestamp, direction, buffer and any later re-entry. Never move A, B or C to erase an already recorded failure.
- The illustrative break rule is not an entry or stop-loss system. A tradable strategy also needs fills, exits, sizing, costs and separate evaluation data.
| Separate bar-23 scenario | Zero buffer | Buffer of 0.5 price points |
|---|
| Low 112; close 113.4 | Wick below; no close-based lower break | No lower break |
| Close exactly 113 | Boundary touch; no break | No lower break |
| Close 112.8 | Lower break | No break: threshold is 112.5 |
| Close exactly 112.5 | Lower break | Threshold touch; no break |
| Close 112.4 | Lower break | Lower break |
Know when the geometry is unusable
Anchor selection is the main discretionary input. Several visible swings can produce different channels on the same chart. A fixed pivot rule makes the selection reproducible, but its lookback and replacement policy still affect the result. Record those choices before judging the fit.
- Require three distinct, chronologically ordered and alternating pivots. Reject a zero horizontal distance from the median origin to M rather than dividing by zero.
- If all three points lie on the same line, the B and C parallels coincide with the median. A zero-width fork provides no channel interior for the break exercise.
- Missing history can prevent confirmation. Leave the fork unavailable rather than filling an unseen pivot or using a later swing retrospectively.
- The arithmetic here assumes linear price coordinates. A logarithmic construction uses transformed prices and needs its own calculation; do not apply these raw-price levels to a log fork. Keep adjusted-price and session settings consistent too.
- A fork selected because it already fits the entire visible chart has hindsight bias. Hide later bars, freeze the available anchors, then record every subsequent outcome, including immediate failures and forks price never revisits.
Build a chart aid and preserve the original hypothesis
Give Pineify Pine Script AI Coding Agent the anchor selection rule, confirmation delay, variant formula and break definition. Ask for a chart aid that displays the three coordinates and marks the confirmation bar separately from the pivot bar. Inspect its source and compare its lines with the worked example before enabling any alerts.
- Specify that a close-based alert requires a completed bar and an active, confirmed fork. New anchors must create a new fork record rather than rewrite earlier signal availability.
- Use Trading Journal to save the plan and later observations. Save the chart screenshot separately and put its share link or file reference in the notes; this preserves a reference to the original view.
- Record symbol, interval, scale, session, A/B/C coordinates, confirmation times, variant, buffer, first break, later re-entry and reason for retiring the fork. Keep the planned rule separate from the outcome review.
- Review the zero-width, missing-history, tied-pivot and exact-threshold cases before relying on generated output. Correct drawing geometry does not establish trading profitability.
