Method coverage map — academy → executable spec
Each rule from academy-notes → its implementation status + where it lives in the technical spec. Serves as the dev's exact todo list, and shows the method is already ~80% specified.
Legend: ✅ executable (quantified/deterministic, in the spec) · ⚙️ partial (defined, threshold/calibration to set) · ⚠️ to code (rule still in prose) · 🧠 judgment (stays with the agent/human, by design).
| Academy rule (§) | Status | Where / what |
|---|---|---|
| Dark pool print = directionless block, soft S/R (§0) | ✅ | data levels; interaction darkpool×node (spec §5) |
| Charts First: fresh > tested (§1) | ✅ | feature fresh = 1 / 0.5 / 0 (§5) |
| "Trade extremes, avoid midpoints" (§1, §6) | ✅ | gate midpoint → PASS (§6) |
| "Good level" (multi-TF, visible reaction, HTF first) (§1) | ⚠️ | multi-timeframe chart logic to code |
| +gamma / −gamma → suppress / amplify (§2,§4) | ✅ | regime from GEX sign |
| ABSOLUTE VALUE RULE (magnitude > color) (§2) | ⚙️ | encoded via node_quality (share of |gex|); magnitude>regime>pattern hierarchy = "expert guess" weights, to recalibrate (§5/§7) |
| King Node = max |exposure| (§3) | ✅ | node_quality (§5) |
| Floor / Ceiling = biggest node below / above spot (§3) | ✅ | derived from the map |
| Gatekeeper = small node between 2 big ones (§3) | ⚠️ | detection to code |
| Air pocket = Σ|gex| < threshold (§3,§8) | ⚠️ | threshold not quantified — to set |
| Day types (range / trend / whipsaw) (§4) | ⚙️ | signatures described → classifier to code |
| Patterns: rug / reverse-rug / pika / beach-ball / whipsaw / rainbow (§5) | ⚠️ | #1 to lock down — pattern weighs ±1.0 in the score (§5) but detection = prose |
| Entry on direct tap, deflection zones ±0.50 QQQ / ±5 SPX (§6) | ✅ | deterministic |
| Stop = 1 node beyond · node-to-node targets (§6) | ✅ | deterministic |
| Node Tap Probability 80 / 66 / 33 (§6) | ✅ | aligned with fresh |
| RR 3:1 / 2:1 / no (§6) | ✅ | gate rr < 2.0 → PASS (§6) + feature rr |
| Regime → execution (absorb the overshoot in −gamma) (§6) | 🧠 | fine timing = agent/human judgment |
| Rolling floors/ceilings = positioning events (§7) | ⚙️ | week-over-week node move detection to code |
| Gamma flip / pivot = directional line (§7, playbook) | ✅ | derived from GEX (zero-line) |
| RoC / Velocity = growing ΔOI (§8) | ✅ | features oi_build / real_node |
| Lifecycle fresh→tested→delivered→decaying (§9) | ⚙️ | fresh coded; "decaying" threshold to define |
| Real vs Hedge node (ΔOI>0 = real) (§9) | ✅ | feature real_node + hedge-only gate (§6) |
| Stairstepping = trend formation (§9) | ⚠️ | detection to code |
| Cross-index / Trinity (SPX·SPY·QQQ) (§10) | ⚙️ | feature index_support; full/partial/divergence logic to code |
| GEX × VEX aligned = booster (supplements) | ✅ | confluence term |
| VIX bands ≤16 / 17-24 / ≥25 (supplements) | ✅ | quantified thresholds |
| Contango / Backwardation = risk-on / off (supplements) | ✅ | macro regime gate (§6) |
| 5-sec playbook (long / short / pivot) (supplements) | ⚙️ | clear rules → setup templates to code |
| Orchestration (Process §11, A+ §12) | 🧠 | the agent's REASON loop + the §5 score |
Summary
- ✅ executable now: ~55% (King/floor/ceiling, fresh, RR, deflection, real-vs-hedge, gamma flip, VIX, contango, GEX×VEX…).
- ⚙️ defined, threshold/calibration to set: ~25% (magnitude-hierarchy, day types, rolling, lifecycle decay, trinity, playbook).
- ⚠️ to code (prose → deterministic): ~15% — the dev's 5 priority workstreams (below).
- 🧠 judgment (by design): fine orchestration + execution timing stay with the agent/human (intended: the edge = judgment on orthogonal inputs, see the AI-human architecture).
The dev's 5 "to code" workstreams (by priority)
- Pattern detection (rug / reverse-rug / pika / beach-ball / whipsaw / rainbow) — rules on the layout of gamma signs around spot. The most important:
patternweighs ±1.0 in the score. - Air pocket threshold (
Σ|gex| < threshold) + node lifecycle "decaying" threshold. - "Day type" classifier (range / trend / whipsaw) from the map signature.
- Rolling / stairstepping detection (week-over-week node moves).
- Trinity logic (full / partial 2-3 / divergence) for
index_support.
None of these 5 is a conceptual blocker — they are rules Léo knows, to be formalized with him into quantified thresholds, then validated in the gauntlet (§7). That is exactly the dev's first task.