Replicate the record yourself

Every statistic on the Validation Lab re-derives from published, derived data — on your machine, with no account, no API key, and (for the standalone path) no installed dependencies. This page is the exact procedure, the exact pass criteria, and the public log of reported replications.

Disclaimer — Research & education only. Not investment advice. Replication verifies reproducibility of published statistics — it is not an endorsement of any signal.

1 · How to reproduce — two equivalent paths

Path A — packaged (pip):

pip install yuclaw
yuclaw replay-lab            # fetches the published bundle and verifies it

Path B — standalone (Python ≥ 3.10 standard library only, no installs):

curl -sO https://yuclawlab.github.io/yuclaw-brain/replay/lab_replay_bundle.json
curl -sO https://raw.githubusercontent.com/YuClawLab/yuclaw-brain/main/tools/replay_lab.py
python3 replay_lab.py lab_replay_bundle.json

Both paths rebuild the decile cohorts from bundled scores, recompute every statistic (bootstrap CI with the published seed, Fama-MacBeth IC with Newey-West correction, market-model alpha), and recompute every sha-256 leaf and daily root against the public yuclaw-trust ledger. Verifier vocabulary: a day counts as exact when the recomputed daily root matches the anchored root byte-for-byte; it counts as an anchored-subset day when the anchored block committed a SUBSET of that day's entry hashes (an intraday re-run anchored before the day completed) and every anchored hash recomputes unchanged from today's data — fewer hashes anchored, none mutated. The replay output names both counts separately; a subset day is a disclosed anchoring artifact, not a verification failure.

2 · What counts as a successful replication

Exit code 0. Any statistic or hash mismatch exits non-zero with a diff report.
Statistics match the bundle's published expected block exactly (same seed, same estimators).
Evidence-ledger roots match — every recomputed daily root equals the evidence-ledger root committed to the public yuclaw-trust repository.
Build metadata matches — the bundle's data-through date, build date, and source commit agree with the page and packet you downloaded them from.

3 · How to report — pass or fail, both are wanted

File a Replication report issue using the template (.github/ISSUE_TEMPLATE/replication.md). Required fields: OS · Python version · exact command · bundle build metadata · sha-256 of the output · result (PASS / FAIL with the mismatch report verbatim). Reports are added to the public log below, unedited.

4 · Public replication log

Entries are recorded verbatim — from filed replication issues, or from the project's own documented external-machine runs — and every entry names its operator and affiliation. Pass and fail alike; never edited retroactively. The disclosure line above the table is derived from the entries: this log never states more independence than the entries record.

External-machine reproduction completed by an affiliated operator; unaffiliated replications: 0

Date	Operator · affiliation	Machine	Source · package	Bundle	Path	Result	Detail
2026-09-04	Claude — project drafter/verifier role · AFFILIATED	external	fresh public clone of YuClawLab/yuclaw-brain at 57a52ec0 · yuclaw from PyPI, fresh venv	`docs/replay/lab_replay_bundle.json sha256 f431f9c629ac38b5… (built 2026-09-01, source 6a63cfb6, ledger e42dd233612a)`	documented manual bundle path (default make-replicate fetch of yuclawlab.github.io blocked by the operator's egress proxy, 403; tool printed the fallback, fallback used)	REPRODUCED	74 daily roots exact + 1 anchored-subset day, 6165 leaf hashes recomputed, all forward (n=70) and in-sample (n=13) statistics match

Status — proven · not proven · accruing

Rendered from one shared source (v3/web/useful_blocks.py) on every page that shows it, so the copies cannot drift. Statuses are measured, not aspirational.

Proven (verifiable today)

✓ Replay works — one command reproduces every Lab statistic and evidence-ledger root from published data
✓ Ledger anchored daily — sha-256 daily roots committed to a public git repository before pages update
✓ Evidence traces to filings — every accepted event carries a source URL, accession number, and verified excerpt
✓ Coverage measured — SEC-filer weight per lens is stated as measured, never rounded up
✓ Snapshots are point-in-time — daily as-of writes, zero retroactive edits (outage window disclosed, not repaired)
✓ Evidence-tier names are never scored — enforced by positive gating and a standing negative check

Not proven

✗ Forward alpha — no spread, IC, or alpha significant at 5% with adequate power
✗ C6 risk-gate sign — rareness confirmed OOS 2026-07-06 (22% fire rate, n=9 held-out); sign confirmation pending (elevated arm n=2; accrual live from 2026-07-16)
✗ Peer-model CAR lead — event-study lead over peer models is not established; live-era sample remains small

Accruing

· Forward out-of-sample record — one period per trading day, accruing daily
· Matured CAR events — each accepted event matures into the event study after its forward window completes
· C6 elevated arm — live Form-4 ingestion since 2026-07-16 restores the insider stream to production inputs
· External replications — the replication log accrues as independent runs are reported

YUCLAW Independent Replication · source commit e124bdffd954 · YuClawLab · research & education only