Trial accounting
Trial accounting is one of 21 measurements this engine publishes in full. It is regenerated from a real run rather than transcribed, and it is shown here with its own claim boundary so the number and its limits arrive together.
The measurement
- Schema
- glassbox.trial-ledger/2
- Immutable execution records
- 231
- Distinct config hashes
- 231
- Window only remeasurements
- 69
- Cross profile duplicate identities
- 0
- Distinct hypothesis identities
- 162
- Ledger profiles
- 4
- Hypothesis identity budget
- 320
- Budget remaining
- 158
- Budget status
- PASS
- Research status
- ACTIVE
- Policy source path
- config/trial_accounting.json
- Generated at
- 2026-08-22T01:05:33.920909+00:00
- Content hash
- sha256:a3b303540e04a477fecc994733e010e231e44e61f7db80f3736531eb10d4224e
Accounting equation. immutable execution records - window-only remeasurements - cross-profile duplicate identities = distinct union hypothesis identities
Selection statistics
- Unit
- first_immutable_record_per_hypothesis
- N hypotheses
- 162
- Sharpe variance
- 9.606e-4
- Audit raw record sharpe variance
- 7.298e-4
Interpretation. Operational window remeasurements remain public but cannot alter selection N or selection V[SR].
Definitions
Immutable execution record. One append-only, config-hash-unique measurement row. Rolling-window remeasurements remain records.
Hypothesis identity. A distinct economic or parameter identity after removing only start and end measurement-window keys.
Window only remeasurement. A later record whose config differs from an existing identity only in start or end. It remains public but does not inflate DSR selection N.
Budget review
- Authorized by
- owner, 2026-08-21
- Hypothesis identities at review
- 162
- Prior budget
- 160
- New budget
- 320
- Headroom
- 158
Authorization. Explicit owner decision to fund the fourteen-sleeve programme, taken with the evidence cost on the table rather than inferred afterwards.
Rationale. Ten new sleeves at the measured 6.5% hit rate (3 survivors from 46 tested candidates) implies about 154 trials. The budget is set at 320 identities, which funds that with a small margin and stops well short of an open-ended search. It is a CEILING, not a plan: the campaign closes at whatever count the evidence stops justifying.
Cost of this decision
- Evidence
- artifacts/analysis/trial_budget_cost/result.json
- Annualized deflated benchmark before
- 1.207400
- Annualized deflated benchmark after 154 trials
- 1.304100
- Increase
- 0.096700
Applies to. every sleeve in the book, not only the new candidates. Deflation is a property of the search, and the search is the book's.
Why it is affordable. The deflated benchmark grows with the LOGARITHM of the trial count, so 154 trials on top of 162 move it by roughly 0.10 annualized. Sample length is the far stronger lever, which is why v6 raising minimum_oos_observations from 252 to 756 matters more to admissibility than this budget does.
The honest caveat. This assumes the measured 6.5% hit rate holds for families that have never been tested. It is an assumption, not a measurement, and the atlas families are deliberately unlike the ones that produced that rate. If the true rate is half of it, ten sleeves cost twice the trials and the budget will bind before the target is reached. Reaching the ceiling without ten admissions is a RESULT, not a failure of the budget, and the book stays smaller.
What did not change. A target still never creates a sleeve. Fourteen is reached only if ten or more new identities independently clear every admission gate in config/sleeve_admission_contract.json; otherwise the book remains smaller. The budget buys permission to look, not permission to admit.
Next tripwire. At 320 recorded hypothesis identities, registration pauses again and requires a fresh owner review. Any single family exceeding 40 identities also pauses for review, so the budget cannot be consumed by one line of enquiry that is not working.
Status token note. ACTIVE is the canonical token the trial tripwire recognises. The prospective nature of this budget is recorded here rather than encoded in the status, so no guard is left unable to read the state it governs.
| Profile | Source path | Immutable execution records | Distinct config hashes | Hypothesis identities | Window only remeasurements | Latest recorded at |
|---|---|---|---|---|---|---|
| var | var/experiments.jsonl | 172 | 172 | 136 | 36 | 2026-08-17T07:52:42.849000+00:00 |
| var_fut_real | var_fut_real/experiments.jsonl | 1 | 1 | 1 | 0 | 2026-06-28T20:13:53.117000+00:00 |
| var_mf | var_mf/experiments.jsonl | 46 | 46 | 13 | 33 | 2026-08-21T05:30:30.254000+00:00 |
| var_sharadar | var_sharadar/experiments.jsonl | 12 | 12 | 12 | 0 | 2026-07-11T19:41:42.107000+00:00 |
| Hypothesis key | Config hash | Label | Ledger profile | First recorded at | Observations | Annualized sharpe observed |
|---|---|---|---|---|---|---|
| eef46f796cdbb646 | 34067635abce9cf9 | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.023000+00:00 | 5428 | -0.018761 |
| 5dd5372deac0882f | d7561b05e7624193 | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.022000+00:00 | 5428 | -0.012095 |
| 1521380bbfea6b18 | 15b78eaf0b3d4820 | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.021000+00:00 | 5428 | -0.054936 |
| 944daf4d5888647d | 103beae582dee892 | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.020000+00:00 | 5428 | -0.030261 |
| 1700ddbeaf47d3eb | 069c86acc068a4d5 | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.019000+00:00 | 5428 | 0.066016 |
| 2cfbbeea11dd0e2e | 3cc8b915abefa48a | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.018000+00:00 | 5428 | -0.292825 |
| 3096d43484ea2745 | 57fa40c3ea31b530 | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.017000+00:00 | 5428 | 0.014885 |
| 8e0d269423afaeb7 | 9584c5e969d4b39a | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.016000+00:00 | 5428 | -0.002788 |
| f350a08ec0da47d5 | ff8fefa17679b51b | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.015000+00:00 | 5428 | -0.053703 |
| 0c17b616473d3e37 | b332501af87f95b6 | forensic_alphamax_weighting | var | 2026-08-17T00:00:00.014000+00:00 | 5428 | -0.050382 |
Trial debt reconciliation
- Source path
- artifacts/audit/trial_debt_reconciliation.json
- Source sha256
- 5ff87cc863c7ce1301f95798d82feb9b5486332eb8a1b2a58258967e14b34630
Legacy dsr debt
- Status
- CODE_RESOLVED_HISTORICAL_CLAIMS_RETIRED
- Historical exception paths
- 12
- Executable debt paths
- 0
- Resolved code paths
- 12
- Union registration paths
- 7
- Source path
- config/legacy_dsr_exceptions.json
- Source sha256
- df0fdfef765f5933ea90e89d131b70ce120c486a792278d07cbbe7326a364e36
Restatement
- Source path
- artifacts/audit/legacy_dsr_restatement.json
- Source sha256
- 3e634254e99ecb8e29eb4f0d807b19e29d54a05f2fc66e2e4493c4a88b49a202
Summary
5 further fields. Read them in the artifact.
Check it yourself
Every figure above is read from /glassbox/trial_accounting.json, the
artifact the engine wrote. Nothing on this page is typed by hand: the page is generated from that
file, so a figure that moves in the artifact moves here and a figure that is not in the artifact
cannot appear here at all.