Loading…
Loading…
162 upcoming measures currently flagged with a date · 6 have since matched an enacted register action (resolved pairs) · gate at 5.
A third calibration channel, distinct from the prediction-market harness: instead of scoring us against an external market, it scores our OWN forward axis-2 rubric against our OWN register. Every time we flag an upcoming measure at a legislative stage (→ an honest likelihood band) and it later becomes an enacted register action, that flagged→enacted pair is a ground-truth resolution of our forward call — sourced entirely from the proprietary, git-versioned action corpus. It answers a buyer's sharpest question, “is your ‘if this passes’ band actually predictive?”, with our own track record.
Separate track. This is the Tier-1 OFFICIAL register self-calibration — never merged with the alt/OSINT prediction-market store or the policy-pressure number. Each resolved pair is matched conservatively (a single unambiguous enacted action by the same issuer, on an overlapping material, announced on/after the flag date and within 18 months) and committed to data/forecasting/self_calibration.json — a reviewable git diff, so the linkage is auditable, not asserted.
Early-warning window over 6 resolved pairs — the lead time between when we flagged a measure and when it became law:
Upcoming measures currently flagged with a date, grouped by the band their legislative stage derives to. This is the denominator the enactment-rate profile above will score once these resolve.
Built only from the proprietary action register + the upcoming-threats queue — no external source, no cron, no fabricated probability. This is the no-regret Rung-1 foundation of the self-calibration channel: it measures whether our forward bands are predictive against our own ground truth; it accrues automatically as flagged measures enact.