What we can and cannot yet say about the MoonshotScore

Engine revision v3.3

600 selected of 125605 eligible records in the date window; 125005 outside the bounded sample.

10 mature member-windows lack matching stock or benchmark prices. These gaps are not zero returns or evidence of delisting; results are withheld.

Measurement protocol

Version 2026-09-28-v4. Protocol revision; not a claim of prospective preregistration for older scores.

• Recorded MoonshotScore V2 scores for common stocks and ADRs.
• Separate known engine revisions; earliest 24 eligible dates in each. Within each revision use oldest date, SHA256 of ticker and date, then ticker. Allocate at most 600 records total by round-robin in first-date/revision order. A changed sample digest is a new cohort edition, not a like-for-like performance trend. This is a bounded sample, not the full scored universe.
• First scheduled NYSE trading session strictly after the UTC score capture day; that exact closing date must exist for both stock and SPY. Missing quotes do not change the date.
• Last scheduled NYSE trading session on or before entry plus 30, 91, 182 or 365 calendar days. Both stock and SPY must have that exact date. Unscheduled closures require a calendar update.
• FMP-only historical-price-eod/full closes from one series vintage per security. Price returns only; no dividend, transaction-cost or slippage adjustment. Provider revisions remain possible.
• Missing mature stock or benchmark prices remain missing, never zero, carried-forward returns or inferred delistings. Such gaps hold publication.
• Each stock return is paired with SPY on the same entry and exit dates. Excess is the median of those paired differences. Displayed SPY return is the median matched benchmark return.
• Known engine revisions have separate results and publication checks; no pooled cross-revision result. Unknown revisions remain in a disclosed held inventory. Historical scores are never recalculated. Dated report evidence is retained for 400 days.
• At least 20 distinct companies with measured outcomes in one horizon; this is a display floor, not statistical significance. Four horizons and all score bands remain visible.
• Historical descriptive association, not a trading strategy, total return, causal effect or proof of predictive power.

Download protocol (JSON) · Measurement status and evidence (JSON) · Dated report history