{"filename":"agent_20260516_1744.md","content":"# Agent Report — Drift Residual Suppresses Controls but Leaves Low Power\n**Date**: 2026-05-16 17:44\n**Piano**: 11\n**Tension explored**: T8_DRIFT_RESIDUAL_REQUIRED\n\n## Claim Under Test\n\nThe paired vol/shock residual from cycle `20260516_1730` failed because `drift_const_vol` still outperformed the intended oriented variants. A drift-residual score might repair the detector if it subtracts matched smooth drift geometry in addition to `vol_only` and `shock_only`.\n\nThe tested score was:\n\n`drift_residual_profile(split) = max(raw_target(split) - max(raw_drift_const_vol(split), raw_vol_only(split), raw_shock_only(split)), 0)`\n\nThe same residual operation was applied to every surrogate before the scan-aware null comparison.\n\n## Question\n\nDoes subtracting matched smooth drift geometry recover planted orientation better than `drift_const_vol` while preserving component-control rejection?\n\n## Experiment Design\n\nScript executed:\n\n`python3 /opt/D-ND_LAB/data/finance/experiments/drift_residual_orientation_20260516_1744.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- Control ceiling: `max(drift_const_vol, vol_only, shock_only)` inside the same seed/split family.\n- Planted splits: 0.35, 0.50, 0.65.\n- Scan grid: 0.20-0.80 in 13 steps.\n- Null baseline: drift-residual profiles against 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 regime persistence; Bai-Perron would test structural breaks; RV-based methods would rank volatility regimes. This cycle is narrower: it asks whether the D-ND orientation object contains signal beyond matched smooth drift, volatility, and transition-shock controls.\n\n## Results\n\n| Variant | Cases | Robust all-null | iid+block5 | iid cluster hit | RV hit | VaR hit | Median drift residual max | Median iid z | Max iid z |\n|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|\n| oriented_full | 36 | 4/36 = 11.1% | 22.2% | 27.8% | 41.7% | 55.6% | 1.262e-05 | -0.176 | 22.386 |\n| oriented_no_shock | 36 | 3/36 = 8.3% | 8.3% | 13.9% | 38.9% | 58.3% | 8.907e-06 | -0.196 | 11.904 |\n| drift_const_vol | 36 | 0/36 = 0.0% | 0.0% | 0.0% | 11.1% | 22.2% | 0.000e+00 | 0.000 | 0.000 |\n| shock_only | 36 | 0/36 = 0.0% | 0.0% | 0.0% | 41.7% | 52.8% | 0.000e+00 | 0.000 | 0.000 |\n| vol_only | 36 | 0/36 = 0.0% | 0.0% | 0.0% | 41.7% | 52.8% | 0.000e+00 | 0.000 | 0.000 |\n\nAggregate:\n\n- `oriented_full_robust_all_null_rate`: 11.1%\n- `oriented_no_shock_robust_all_null_rate`: 8.3%\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`: +11.1%\n- `delta_no_shock_vs_drift_const_vol`: +8.3%\n- `delta_full_vs_vol_only`: +11.1%\n\nBest robust examples:\n\n- Best `oriented_full`: seed 6201, split 0.65, iid/block5/block21 z = 9.707 / 26.375 / 31.911. Hits planted split.\n- Best `oriented_no_shock`: seed 6201, split 0.50, iid/block5/block21 z = 11.904 / 3.566 / 12.934. Hits planted split.\n- Strongest non-robust `oriented_full`: seed 6205, split 0.50, iid/block5/block21 z = 22.386 / 12.823 / -0.183. Block21 rejects.\n\n## Key Findings\n\n1. The drift-residual ceiling drives robust all-null passes for the three component controls to 0/36 = 0.0%. `drift_const_vol`, `shock_only`, and `vol_only` still retain non-zero raw/RV/VaR signals, but they do not pass the robust all-null gate. This repairs the specific false-positive pathway from the previous cycle.\n\n2. The repair is still not promotable. `oriented_full` remains at 4/36 = 11.1% recovery, and `oriented_no_shock` drops to 3/36 = 8.3%. Both are far below the >=70% calibration target.\n\n3. The positive signal is now specific but sparse. `delta_full_vs_drift_const_vol` becomes positive (+11.1%) because the drift control is zeroed, but the oriented recovery is unchanged from the paired residual cycle. This means the detector gained component-control rejection at 0/36 = 0.0% robust all-null passes, not power.\n\n4. Block21 remains the hard null. Several high iid/block5 cases fail under 21-day circular block shuffle; the strongest non-robust `oriented_full` case has iid z = 22.386 and block5 z = 12.823, but block21 z = -0.183.\n\n5. VaR remains the stronger locator on planted oriented variants. `oriented_full` VaR hit = 55.6% and `oriented_no_shock` VaR hit = 58.3%, while iid cluster hit is 27.8% and 13.9%.\n\n## Verdict\n\n**NO_DELTA (drift-residual score rejected as promotion rule; accepted as a purity control).**\n\nDo not promote to real-market testing. The drift-residual statistic suppresses `drift_const_vol`, `vol_only`, and `shock_only` to 0.0%, but it leaves planted oriented recovery at only 11.1% / 8.3%. The next valid move is not another component ceiling; it is a power-focused redesign that preserves the successful control rejection while increasing recovery on planted oriented transitions.\n\n## Bicono della scoperta\n\n- **Due radici**:\n  - Root 1: Adding `drift_const_vol` to the residual ceiling removes the last explicit component false positive.\n  - Root 2: Once drift-only geometry is subtracted, very little oriented signal remains recoverable by the current scan-cluster object.\n- **Singolare**: The singular point is the split between purity and power: controls are 0/36, but `oriented_full` is still only 4/36.\n- **Invariante di passaggio**: The invariant is no longer \"subtract more controls\"; it is \"retain component-control rejection while increasing planted oriented recovery under iid + block nulls.\"\n- **Campo di possibilita'**: Next cycle should test a power-preserving statistic, such as signed lag-map circulation or local antisymmetric covariance normalized by drift-control energy, with the current drift/vol/shock residual kept as a mandatory negative-control gate.\n\n## Files\n\n- Experiment script: `data/finance/experiments/drift_residual_orientation_20260516_1744.py`\n- Experiment output: `data/finance/experiments/drift_residual_orientation_20260516_1744.json`\n- Report: `data/finance/reports/agent_20260516_1744.md`\n- Seed updated: `data/finance/seed.json`\n","title":"Agent Report — Drift Residual Suppresses Controls but Leaves Low Power","verdict":"NO_DELTA (drift-residual score rejected as promotion rule; accepted as a purity control).","bicono":{"roots":"- Root 1: Adding `drift_const_vol` to the residual ceiling removes the last explicit component false positive.\n  - Root 2: Once drift-only geometry is subtracted, very little oriented signal remains recoverable by the current scan-cluster object.","singular":"The singular point is the split between purity and power: controls are 0/36, but `oriented_full` is still only 4/36.","invariant":"The invariant is no longer \"subtract more controls\"; it is \"retain component-control rejection while increasing planted oriented recovery under iid + block nulls.\"","field":"Next cycle should test a power-preserving statistic, such as signed lag-map circulation or local antisymmetric covariance normalized by drift-control energy, with the current drift/vol/shock residual kept as a mandatory negative-control gate."},"size":6476,"mtime":"2026-05-16T17:50:02.590996+00:00"}