{"filename":"agent_20260516_1759.md","content":"# Agent Report — Block21 Short-Lag Orientation Has No Recoverable Power\n**Date**: 2026-05-16 17:59\n**Piano**: 13\n**Tension explored**: T9_POWER_AFTER_DRIFT_CONTROL\n\n## Claim Under Test\n\nThe drift-residual score from cycle `20260516_1744` repaired component-control purity but left planted oriented recovery at only 11.1% / 8.3%. The next proposed detector was not another residual ceiling: it tested whether short-lag M orientation survives after removing 21-day smooth drift.\n\nScore under test:\n\n`shortlag_score(split) = |mean(circulation_right) - mean(circulation_left)| / local_residual_energy`\n\nwhere returns are first residualized by subtracting a 21-day rolling mean, and:\n\n`circulation_t = det([r_t, r_{t-1}], M [r_t, r_{t-1}])`, with `M = [[1,1],[1,0]]`.\n\n## Question\n\nDoes short-lag M orientation after removing 21-day smooth drift recover planted orientation under a block21-preserving null while keeping drift/vol/shock controls at zero robust passes?\n\n## Experiment Design\n\nScript executed:\n\n`python3 /opt/D-ND_LAB/data/finance/experiments/block21_shortlag_orientation_20260516_1759.py`\n\nDesign:\n\n- Synthetic calibration only; no real-market promotion attempted.\n- 180 cases: 12 seeds x 3 planted splits x 5 variants.\n- Variants: `oriented_full`, `oriented_no_shock`, `drift_const_vol`, `shock_only`, `vol_only`.\n- Planted splits: 0.35, 0.50, 0.65.\n- Scan grid: 0.20-0.80 in 13 steps.\n- Local M-orientation window: 96 observations around each candidate split.\n- Smooth drift removal: 21-day rolling mean subtraction before short-lag circulation.\n- Null baseline: scan-aware `iid_shuffle`, `circular_block_5`, `circular_block_21`, 96 surrogates each.\n- Robust pass: `DND_DELTA` only if non-endpoint cluster passes every null family with `cluster_effect_z >= 3.0` and `p <= 0.05`.\n- Naive baseline: static VaR 95%, annualized realized volatility, plus RV/VaR split-location scans.\n\nPrior-art boundary: Hamilton/HMM would model latent state persistence, Bai-Perron would test parameter breaks, and RV-based methods would rank risk-scale shifts. This cycle does not claim a market regime; it tests whether a D-ND short-lag orientation object has power beyond smooth drift and block-scale order preservation.\n\n## Results\n\n| Variant | Cases | Robust all-null | iid+block5 | iid cluster hit | RV hit | VaR hit | Median block21 z | Max iid z | Max block21 point z |\n|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|\n| oriented_full | 36 | 0/36 = 0.0% | 0.0% | 16.7% | 41.7% | 55.6% | -0.237 | 6.387 | 3.590 |\n| oriented_no_shock | 36 | 0/36 = 0.0% | 0.0% | 13.9% | 38.9% | 58.3% | -0.248 | 10.669 | 4.024 |\n| drift_const_vol | 36 | 0/36 = 0.0% | 0.0% | 0.0% | 11.1% | 22.2% | -0.230 | -0.116 | 3.628 |\n| shock_only | 36 | 0/36 = 0.0% | 0.0% | 11.1% | 41.7% | 52.8% | -0.243 | 6.606 | 3.306 |\n| vol_only | 36 | 0/36 = 0.0% | 0.0% | 8.3% | 41.7% | 52.8% | -0.245 | 8.949 | 3.720 |\n\nAggregate:\n\n- `oriented_full_robust_all_null_rate`: 0.0%\n- `oriented_no_shock_robust_all_null_rate`: 0.0%\n- `drift_const_vol_false_positive_rate`: 0.0%\n- `shock_only_false_positive_rate`: 0.0%\n- `vol_only_false_positive_rate`: 0.0%\n- `delta_full_vs_drift_const_vol`: 0.0%\n- `delta_no_shock_vs_drift_const_vol`: 0.0%\n- `delta_full_vs_vol_only`: 0.0%\n\nBest non-robust examples:\n\n- Best `oriented_full`: seed 6201, split 0.65, iid z = 6.387, p = 0.010, planted split hit = true; block5 and block21 both reject, block21 z = -0.215.\n- Best `oriented_no_shock`: seed 6201, split 0.65, iid z = 10.669, p = 0.010, planted split hit = true; block5 and block21 both reject, block21 z = -0.279.\n- Best control leakage remains visible before block nulls: `shock_only` iid z = 6.606 and `vol_only` iid z = 8.949, both with planted split hits, but neither survives block5/block21.\n\n## Key Findings\n\n1. The block21 short-lag object has no robust planted recovery. `oriented_full` and `oriented_no_shock` both fall to 0/36 robust all-null passes. This is below the previous drift-residual recovery and far below the >=70% calibration target.\n\n2. The object is pure only because block nulls reject everything. Component controls are 0/36 robust passes, but so are planted positives. This is not a usable detector; it is an over-hard null/filter combination.\n\n3. The iid-only signal is not discriminating enough. The best iid positives hit the planted split, but controls also produce high iid z and planted-hit artifacts (`shock_only` z = 6.606, `vol_only` z = 8.949). Block preservation removes them, but also removes the positives.\n\n4. Block21 is again the decisive falsifier. Max block21 point z can exceed 3 on several variants, but cluster-effect z remains negative and p = 1.0 in the strongest inspected block21 cases. Point z without non-endpoint cluster mass is not promotable.\n\n5. VaR remains stronger as a split locator on planted oriented variants: 55.6% / 58.3% VaR hit versus 16.7% / 13.9% iid cluster hit. D-ND still has no detector advantage on this synthetic calibration.\n\n## Verdict\n\n**NO_DELTA. The claim that block21 short-lag orientation could recover planted orientation while keeping controls at 0.0% robust passes was falsified at this cycle: `oriented_full` and `oriented_no_shock` both produced 0/36 = 0.0% robust all-null passes. The new claim that emerged is a negative constraint: block21-survival cannot be treated as the positive detector unless the positive object first shows discriminating recovery under a less destructive paired/block design.**\n\nDo not promote to real-market testing. Removing 21-day smooth drift and measuring short-lag M circulation produces clean all-null rejection, but only by zeroing both positives and controls. The useful constraint is sharper: block21-survival cannot be treated as the positive detector unless the positive object first shows discriminating recovery under a less destructive paired/block design.\n\n## Bicono della scoperta\n\n- **Due radici**:\n  - Root 1: Short-lag M circulation can produce iid split hits on planted positives.\n  - Root 2: The same iid-level behavior also appears in `shock_only` and `vol_only`, so it is not a clean orientation signature.\n- **Singolare**: The singular point is block21 collapse: all variants are 0/36 robust all-null, including the intended positives.\n- **Invariante di passaggio**: A promotable detector must first separate positives from component controls before block21 is used as the final robustness gate.\n- **Campo di possibilita'**: Next cycle should stop testing single hard detectors against block21. Test a two-stage design instead: use a higher-power candidate to recover planted positives under iid/block5, then apply drift-residual/block21 as a veto, not as the score itself.\n\n## Files\n\n- Experiment script: `data/finance/experiments/block21_shortlag_orientation_20260516_1759.py`\n- Experiment output: `data/finance/experiments/block21_shortlag_orientation_20260516_1759.json`\n- Report: `data/finance/reports/agent_20260516_1759.md`\n- Seed updated: `data/finance/seed.json`\n","title":"Agent Report — Block21 Short-Lag Orientation Has No Recoverable Power","verdict":"NO_DELTA. The claim that block21 short-lag orientation could recover planted orientation while keeping controls at 0.0% robust passes was falsified at this cycle: `oriented_full` and `oriented_no_shoc","bicono":{"roots":"- Root 1: Short-lag M circulation can produce iid split hits on planted positives.\n  - Root 2: The same iid-level behavior also appears in `shock_only` and `vol_only`, so it is not a clean orientation signature.","singular":"The singular point is block21 collapse: all variants are 0/36 robust all-null, including the intended positives.","invariant":"A promotable detector must first separate positives from component controls before block21 is used as the final robustness gate.","field":"Next cycle should stop testing single hard detectors against block21. Test a two-stage design instead: use a higher-power candidate to recover planted positives under iid/block5, then apply drift-residual/block21 as a veto, not as the score itself."},"size":6969,"mtime":"2026-05-16T18:07:56.011675+00:00"}