Fractal Model Desk fractal model indicator for TradingView

Five-minute charts automatically use one hour. All pairings appear in Settings below.

Use standard time-based candles. The larger interval must exceed the chart interval, with each chart candle wholly inside it. Boundary-crossing candles raise an error. The first partial larger interval is skipped.

C1 is the previous closed larger candle. A wick strictly beyond its low or high records a raw C2 candidate. No reclaim is required. One candidate per direction per larger candle is retained, including both on a dual sweep.

C2 closure needs its final close strictly inside the whole C1 range. Equality fails. Closure and CISD remain separate raw facts; neither alone qualifies for delivery review.

A larger close becomes known at the next interval’s first confirmed small-candle close. On 5m / 1h, C2 ending at 10:00 is recorded at 10:05, or later after a session gap.

Markers use observation time. In Audit, the dotted C1 reference points back to its source candle; Focus starts it no earlier than review. Atlas FORMING is a live preview, not confirmation; the Lab preview updates at chart-bar close.

The opposing candle series behind the extreme supplies its opening-price line. A confirmed small candle closes strictly beyond it. Dojis end a series; only a new strict C2 extreme can update a pending anchor.

First CISD freezes the line, extreme and projection leg. Its delivery quality is also measured then using prior volatility. A later strong candle cannot improve weak or missing evidence. A strict extreme breach records failure without erasing the earlier observation.

After failed C2 closure, bullish C3 can still hold C2's wick low and close above both C2 body prices; bearish mirrors it. This fallback is a declared implementation rule.

C3/C4 name stages, not entries. Strong C3 closes directionally beyond C2's range. C4 completion ends the observation window, not a broker position. New parent CISD cannot first confirm in C4.

After closure, a strict breach of fixed C2 extreme invalidates structure: wick by default, close optionally. Equality survives. Confirmed lines and observation times remain in the record.

A review can become Withdrawn or Complete without losing its original evidence. These are review states, not losses or wins. Focus selects two by original review time; Audit shows five recent raw records.

Raw history defaults to 24, adjustable 8–40. Oldest completed records are evicted first; developing records are protected. IC has a separate 24-record history, with active/current-frame overflow protected.

No. A raw FVG is the first new three-candle gap after known, unfailed CISD, starting no earlier than the tracked extreme. Audit begins its box at detection; Focus starts it no earlier than review. A retest needs a later overlapping candle.

Parent delivery review does not require an FVG. The limited IC-review route does. A gap or overlap proves neither trader intent nor a limit fill.

Closure and CISD must belong to the matching larger candle, with enough prior volatility history. The first CISD needs a decisive close and either a quick, direct V reversal or a strong directional candle.

Repeated overlapping ranges, dual sweeps, failed structure or a spent objective can block review. Thresholds are our adjustable heuristics. A first CISD without 14 prior chart bars is Excluded; later data cannot repair that frozen measurement. A review still leaves POI and bias to you.

IC follows a pullback inside C3/C4 of an already-reviewed parent. It has its own ID, opposing-leg opening level and protected three-candle swing. It never moves the parent anchor.

IC anatomy needs an opposing leg through the current larger-candle open, a later close back through that leg’s open, and a swing already confirmed by its right neighbor. IC review also needs a parent FVG already known when the touch bar opens, plus displacement. A same-bar new gap or later touch cannot improve earlier anatomy.

One IC attempt starts per parent per C3/C4 frame, even without an FVG; its confirmed swing must remain the pullback extreme. Attempts end with the frame or parent review. This does not cover every POI model. A protected-swing breach records a child outcome separately. Focus shows two IC reviews; Audit can show two formed anatomies. The Lab does not trade them.

No. It records a sweep/reclaim of an already-known local pivot before the outer candidate. It neither establishes anyone’s intent nor gates delivery review. IC protected-swing failure is also separate from parent invalidation.

Origin + direction × multiple × frozen leg. The 1 / 2 / 2.5 / 4 labels are price-leg multiples, not standard deviations or promised targets. The 1× line equals CISD.

For CISD 110, wick 100 and opposing-series extreme close 104: Wick gives 110 / 120 / 125 / 140; Body gives 110 / 116 / 119 / 128. Invalid geometry is not rescued with an absolute value.

Latest review can select an ended record. Focus never starts its displayed levels before review became known. An old projection is not a current entry.

Delivery review acts on a new qualifying review. FVG retest after review requires a later retest event with the review already known when that retest bar opens. A first retest before or in the review bar is skipped; no replacement retest is invented.

Both need active review, paired intact structure, unfailed CISD and an available C1 opposite-extreme target. IC children are not traded. One position and one submission per calculation are allowed. These restrictions never hide Atlas observations.

Stop uses fixed C2 extreme plus tick padding. Target uses the opposite C1 extreme; there is no 2R fallback. Spent or behind-entry targets and invalid distances are skipped. The entry submits its stop/limit bracket immediately. C4 completion does not itself close a broker position.

Audit shows neutral raw observations. Match IDs and observation times across the ledger and chart; Times uses the exchange timezone. FORMING is a live preview, never a review.

Focus instead selects the latest two reviews, including withdrawn/completed ones. Display filtering preserves engine history. Latest review projections can therefore belong to an ended review.

Delivery review is the default stream: delivery/IC reviews and later outcomes. All observations also includes neutral raw facts and IC anatomy. Enable JSON batches and create Any alert() function call.

The schema is fm-desk-v2.1 with scope and full_model_validation: false. IDs, observation times and evidence travel with each event. Update old receivers and recreate alerts. Named alert conditions are indicator-only; Lab broker notifications are separate.

The producer's release evidence records 228/228 native Pine assertions across 50 fixtures: core 91, delivery/context 86 and IC 51. Its build/package tests passed 18/18. Both scripts and three probes compiled without errors or warnings.

Market smoke checks, prefix tests and same-history reconstruction have documented limits. No native five-minute Bar Replay or paid-indicator parity pass is claimed. These checks establish neither trading accuracy nor profit. Read the evidence and its limits.

The Lab's 10% equity position size is allocation, not stop-loss risk. Review costs, Properties and emulator fills separately.

POI means the price area you choose to study; bias means your larger directional view. Both remain manual, alongside the objective and trade risk. Review is not full model validation.

Published research informed these disclosed heuristics, not a recovered proprietary formula. Settings, feeds and loaded history can change recalculation. No blanket repaint or paid-indicator parity promise is made. Read the three-project source audit.

Use standard 5m / 1h and default Focus. Read one review's evidence, then compare its raw sequence in Audit. An empty Focus view can be correct. Learn the separate entry and cost rules before experimenting with the Lab.