trd.fun / lab
Archive

Wave-9 programme status — 5 September 2026

Updated 2026-09-05
Superseded programme snapshot: This preserved in-progress trial-61 record is replaced by the completed trial-63 status dated 8 September.
SUPERSEDED
docs/research/wave9-programme-status-2026-09-05.md

Superseded programme snapshot: This preserved in-progress trial-61 record is replaced by the completed trial-63 status dated 8 September.

Status date: 2026-09-05. This is a programme-status snapshot, not a result or an authority record. (Source: /home/sam/gates/W9-STATUS-2-brief.md:6-10.)

Public documentation · paper-only research · not investment advice. Nothing on this page authorizes an allocation, a broker action, or a live order. (Source: app/components/lab-v3/DocsPageShell.tsx:61-63; docs/research/funding-gate-runbook-2026-09-04.md:92-98.)

Registered family

The active registration is trial 61, ledger entry wave9-exploration-sweep-r5. It binds the fixed family of 84 declared rows in 14 clusters to declaration SHA-256 5048eba37eda9048f1e854a308145ba93e1e90c57aef3f0985fa3ca5b0e69558, manifest SHA-256 c8d97bf7d194841156770bf686641ad4f1d614200187bae92f85fe85633131fc, and ledger head 9d2f2dbdc935377ef986fc1bb558d6beee0caf7184fef122d5bcb3cc3591ce01. (Sources: /home/sam/gates/logs/W9-DECL-4.before-r2.final.md:1; lib/live-lane/sweep/index.ts:20-26; scripts/wave9-sweep-publication.mjs:14; scripts/wave9-sweep-run.test.mjs:205.)

The tree registers 66 implemented rows and the manifest records the other 18 as exclusions, with no unresolved row. (Sources: lib/live-lane/sweep/index.ts:29-103; /home/sam/gates/logs/REVIEW-W9-DECL-4.verdict.md:4-5,12.)

An exclusion is a recorded disposition, not a zero return and not permission to replace a row.

Ratified conventions and re-expressed rows

ChangeRegistered effect
Zero-proceeds amendmentA documented terminal event with exactly zero proceeds is a typed total-loss write-off rather than a price observation. The owner-ratified amendment is implemented by PR #402 and was first registered at trial 60. (Source: docs/research/wave9-zero-proceeds-amendment-2026-09-04.md:3-4,19-46,56-59.)
KERN-7PR #417 added opt-in declining select_weights at the gross cap; the default remains refusal. Trial 61 re-expressed W9-002, W9-087 and W9-103 under atCap: 'decline'. (Sources: lib/live-lane/kernel/index.ts:211-214,1083-1091,4217-4232; /home/sam/gates/W9-STATUS-2-brief.md:10,16.)
KERN-8PR #422 implemented the owner-ratified sameBarResolution: 'prior_stop' convention while preserving fail-closed refusal as the default. Trial 61 re-expressed W9-039, W9-043 and W9-048 under it, and each use emits typed receipt evidence. (Source: docs/research/wave9-same-bar-trailing-stop-convention-2026-09-05.md:3-4,8-24,26-31.)
W9-089 correctionThe monthly votes now anchor to the last completed month end less than or equal to the decision date; PR #420 carried the correction into the PR #423 trial-61 train. (Sources: lib/live-lane/strategy/wave9/w9-089-simple-majority-trend-ensemble.ts:50-72,87-99; lib/live-lane/strategy/wave9/w9-089-simple-majority-trend-ensemble.test.ts:47-57; /home/sam/gates/W9-STATUS-2-brief.md:16.)

These conventions and the correction define research semantics. They are not performance findings and grant no allocation or execution authority.

What remains excluded

All 18 exclusions remain in postImplementationCapabilityGap. W9-007 and W9-045 remain there by owner ruling with reasons additive to their earlier recorded gaps; the other 16 require kernel capabilities that are not built. (Sources: /home/sam/gates/W9-STATUS-2-brief.md:16; /home/sam/gates/logs/REVIEW-W9-DECL-4.verdict.md:4.)

The three KERN-7 rows are no longer exclusions: W9-002, W9-087 and W9-103 are present in both the sweep table and the paper registry. (Sources: lib/live-lane/sweep/index.ts:29-31,89-96; lib/live-lane/paper-registry.ts:114-181.)

Compute status

Earlier Stage-1a material remains calibration only; no result computed under superseded pins carries forward. (Sources: /home/sam/gates/W9-STATUS-1-brief.md:10; /home/sam/gates/stage1b-t61.sh:4.)

At trial 60, Phase A sealed 44 rows in three groups of 18, 11 and 15 on two lanes. Two partial Phase-B runs stopped after the shared CI runner was OOM-killed, and no SEP row sealed. Those trial-60 artifacts are trial-bound parity references and refusal-discovery evidence only; 17 fields are required not to move during parity checks. They are not the publishable Wave-9 batch. (Source: /home/sam/gates/W9-STATUS-2-brief.md:10,16.)

The trial-61 full-family recompute is in progress. It uses the versioned launcher in direct rows-only mode, one process per spec; sealed rows cannot be reused across the trial boundary. The authoritative batch can be formed only after the complete registered row set is combined. (Sources: /home/sam/gates/W9-STATUS-2-brief.md:10,16-17; /home/sam/gates/stage1b-t61.sh:2-4,31-36,45-52,64-66.)

No trial-61 recompute receipt exists yet, so this page publishes no progress count and no batch result. (Source: /home/sam/gates/W9-STATUS-2-brief.md:17.)

History

  • Trial 59 was the prior registered baseline. Trial 60 superseded it to register 11 Stage-2 rows under KERN-6 and the ratified zero-proceeds semantics. (Sources: /home/sam/gates/W9-STATUS-1-brief.md:16; /home/sam/gates/receipts/W9-DECL-3.final.md:1; docs/research/wave9-zero-proceeds-amendment-2026-09-04.md:48-59.)
  • Trial 60 registered 63 implemented rows and 21 exclusions, with declaration 9006e6b1…, manifest aa68a4e7… and head ad215de8…; its compute exposed the remaining convention and spec issues. (Sources: /home/sam/gates/receipts/W9-DECL-3.final.md:1; /home/sam/gates/W9-STATUS-2-brief.md:16.)
  • Trial 61 re-registered after KERN-7, the owner-ratified KERN-8 convention and the W9-089 correction, producing the current 66-implemented/18-excluded definition. (Sources: docs/research/wave9-same-bar-trailing-stop-convention-2026-09-05.md:26-31; lib/live-lane/sweep/index.ts:24-26,29-103; /home/sam/gates/W9-STATUS-2-brief.md:16.)

What the paper lane can do today

These are software capabilities and refusals. They do not make a row paper-eligible by themselves.

Delivered packetCurrent capability
W9P-1 (PR #409)An explicit paper-mode selector context relaxes only the 2021-12-31 research cutoff; every other validation and registry authority remains. (Sources: lib/live-lane/strategy/index.ts:85-93; /home/sam/gates/W9-STATUS-1-brief.md:16.)
W9P-2 (PR #397), W9P-2b (PR #406) and W9P-2c (PR #412)Eligibility requires the operator's absolute ledger path, a verified chain and optional head pin, a canonical registered spec, exactly one matching preregistration, paper-only status and a paper broker origin. (Sources: lib/live-lane/paper-registry.ts:224-284; scripts/live-lane.mjs:434-452; git merge commits 8f748396 (#397) and 7e6e2d0a (#406); /home/sam/gates/W9-STATUS-1-brief.md:16 (#412).)
W9P-3 (PR #398) and W9P-3b (PR #408)The session projector mirrors the sealed lot-lifecycle grammar and refuses unsupported execution or sizing; it advances carried lots, exits and entries from causal state. (Sources: lib/live-lane/strategy/lot-projection.ts:252-299,301-369,376-451; git merge commit c8f18701 (#398); /home/sam/gates/W9-STATUS-1-brief.md:16 (#408).)
W9P-4 (PR #416)--strategy W9-0NN selects an admitted strategy, projects a paper plan, submits only after paper/live mode checks, and reconciles against Alpaca PAPER. (Sources: scripts/live-lane.mjs:434-452,685-701,842-919,1155-1175,1462; lib/live-lane/paper-lane.ts:95-131; /home/sam/gates/W9-STATUS-1-brief.md:16.)
W9P-6 (PR #419)Reconciliation is fill-aware: zero, full, partial and ambiguous fills update or refuse hash-chained lot state, and recovery keeps incident-lineage attribution. (Sources: lib/live-lane/paper-lane.ts:220-264,405-489,571-600; scripts/live-lane.mjs:1223-1242,1296-1316; /home/sam/gates/W9-STATUS-2-brief.md:17.)

How paper candidates will be chosen

The pre-registered selection rule grants no paper or live authority. It is applied only after the authoritative combined batch receipt exists. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:3,9-10.)

GateRequirement
S1Reconciliation, exact row census and cryptographic verification must pass. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:11.)
S2Batch PBO must be at most 0.20; otherwise the whole batch selects nothing. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:12.)
S3The row must pass Holm at exploration alpha 0.10; BH is report-only. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:13.)
S4Clustered DSR probability must be at least 0.95; raw DSR is report-only. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:14.)
S5The 25-basis-point-per-side stress endpoint must be positive. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:15.)
S6From 2017-01 onward, 12 times the arithmetic monthly candidate-minus-benchmark mean must be positive. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:16.)
S7Keep at most one row per family cluster, choosing the highest clustered DSR and breaking ties by batch ID. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:17.)

After S1–S7, rank cluster winners by clustered DSR and keep at most 3. An empty list is valid. (Source: docs/research/wave9-paper-selection-rule-2026-09-05.md:17-20,24-28.)

Earliest dates, not promises

The paper-lane design placed account preparation on 21 September 2026 and the first possible paper session on 22 September 2026. That date remains conditional on selection, fresh preregistration and the paper authority chain; it is not authorization to submit. (Source: /home/sam/gates/receipts/ll-w9-design.md:371-385.)

Under the design's three-countable-interval floor, the earliest honest graduation decision was approximately 5 January 2027, conditional on a Wave-9 owner amendment defining statistical validity. Even then, a separate Wave-9-capable graduation path and independent review would be required before any live action. (Source: /home/sam/gates/receipts/ll-w9-design.md:386-395.)

No batch result, selection receipt, paper observation or date on this page authorizes a live order. The standing rule remains: no live order follows from any of this. (Sources: docs/research/wave9-paper-selection-rule-2026-09-05.md:3; docs/research/funding-gate-runbook-2026-09-04.md:92-98.)

On this page 8 sections
LAB documentation describes design intent and research safeguards. It does not provide trading instructions or operational access.