ICT PO3 AI [PickMyTrade]

Every AMD / Power-of-Three script marks where accumulation, manipulation and distribution happened. None answer the question a trader actually has the moment a range locks and gets swept: is THIS raid the kind that expands into distribution, or a pattern that only looked right?

――――――――――――――――――――――――――――――――――――――


🔷 WHAT IT MEASURES

• A locked accumulation range, scored on five independent compression measures — not one pass/fail test
• Every raid checked against a weighted pool of real liquidity — previous week/day high-low, Asia high-low, last confirmed swing — never against price alone
• A reclaim's own quality — penetration depth, acceptance beyond the level, overshoot, rejection wick — scored before the sequence is trusted
• The expansion leg itself: a genuine displacement bar, the first qualifying imbalance, and a market-structure shift back through the range it came from

――――――――――――――――――――――――――――――――――――――

🔷 THE ENGINE

• Three phase scores — Accumulation, Manipulation, Distribution — combined by weighted geometric mean, not a sum: a weak link drags the whole score down, never averaged away
• A k-nearest-neighbour search compares every new setup's own 8-property fingerprint against up to 100 stored past outcomes from the same chart — no training, no forecasting, just distance
• Whether a signal exists at all is decided by measured price behaviour alone; the analog engine sharpens the read but — outside Gate mode — never adds or removes a signal
• Below enough same-direction history, the high-confluence circle tier is disabled outright rather than guessed

――――――――――――――――――――――――――――――――――――――

🔷 SIGNALS AND DISPLAY

• Accumulation and Distribution boxes colored by role, always the same hue for the same tag — never by market bias
• A raid marker (×) placed at the sweep extreme, drawn only once the reclaim actually confirms — a location, not an entry
• Circle = high confluence, triangle = standard signal; marker opacity (not size) scales with score
• Dashboard: phase, the three sub-scores, regime, which liquidity level was raided, entry/stop/target, R-multiple, and the analog library's own sample count
• 6 alertconditions, plus an optional JSON payload for webhook use

――――――――――――――――――――――――――――――――――――――

🔷 INPUTS

• ATR Length / Compression Window — volatility baseline and consolidation lookback. Default 14 / 20.
• Consolidation Sensitivity — Strict/Balanced/Loose compression floor. Default Loose.
• Minimum Score / High Confluence Score — structural floor to draw a signal, and the circle-tier threshold. Default 45 / 80.
• Higher Timeframe / Use Session Context / Timezone — directional context and session-window fit.
• Fallback Target Multiple — used only when no liquidity level qualifies as a target. Default 2.0R.
• Adaptive Mode — Off / Advisory / Gate. Default Advisory.
• Neighbours Compared / Outcome Horizon — KNN neighbour count and bars until a signal's result is recorded. Default 8 / 24.
• Show Accumulation Range, Distribution Zone, Imbalance — display toggles for each zone type.
• Zen Mode, Show Dashboard, Dashboard Position/Detail — display toggles; Zen Mode hides everything but zones and signals.

――――――――――――――――――――――――――――――――――――――

🔷 REQUIREMENTS AND LIMITATIONS

• Swing highs/lows used in the liquidity-pool weighting confirm 10 bars after they form, by construction — a level newer than that isn't counted yet
• The analog engine's read is thin until its library builds; the sample count travels with every score so that's never hidden
• Stored past outcomes are hypothetical — no commission, no slippage — and are not a record of trading results
• This script measures a specific, disclosed sequence. It does not predict, and it is not a trading system on its own.

――――――――――――――――――――――――――――――――――――――

Built in Pine Script v6. Open source — Mozilla Public License 2.0.