Robotics Risk Ledger — track record

A pre-registered, honestly-graded record of humanoid/physical-AI calls: which robot claims were real (deployment truth) and where the picks-and-shovels value accrued. Every call is frozen with a measurable contract before the outcome exists and scored against reality AND a naive baseline. We have not identified an incumbent that publishes this combination (forward-frozen + machine-graded + baseline-compared + externally timestamped).

*In plain English: The public track record — the aggregate scoreboard only, with no live calls or methods shown. It stays deliberately humble: until the engine has graded enough calls AND beaten the do-nothing baseline, this page shows exactly how far it still is from being allowed to claim it works.

★For the kids: This is the robot engine's honest report card — but only the final grades, no peeking at how it works inside. And it won't show off a single grade until it has passed enough tests to prove the score is real and not just luck.

Record structure — read this before the numbers

4 frozen Tier-1 calls across 2 independent thesis families. Effective-N: distinct INDEPENDENT thesis families among the calls. When this is much less than the raw call count, correlated bets are inflating the headline — the record is only as strong as its independent families.

Tier-1 (the spine): Tier-1 = the auditable SPINE: machine-graded from market/financial point-in-time data (proxy-basket returns + XBRL revenue) against a pre-registered contract AND a persistence null. This is the bulletproof record — no judgment in the grade — and the ONLY tier the public credential is 'earned' on.

Tier-2 (supporting): Tier-2 = SUPPORTING evidence: auto-graded from the citation-gated boards + cited-event ledgers (deployment/capability rungs, roadmap slips, advantage survival). Auditable, but its inputs are operator/crew judgment — reported separately and NEVER counted toward the earned threshold.

Accumulating — credential not yet earned

Tier-1 calls graded0 / >= 20

Accumulating — 0 of 20 Tier-1 calls graded. The scoreboard publishes at >=20 graded eligible outcomes, WIN OR LOSE — publication is unconditional (SCOREBOARD_PROTOCOL.md v1, frozen 2026-07-22 before any outcome existed); only the earned CREDENTIAL separately requires demonstrated skill. Freeze-means-freeze.

Aggregate calibration only; not investment advice, and NOT a list of robot stocks. The engine's live calls + methodology are private. · as of 2026-09-09