ResetAll history; 6 in this period
Verified history first. Facts, estimates and unknowns kept distinct.
7 eligible events, 0 with comparable scope and exact timing.
All history; 6 in this period
6 comparable intervals, not a fixed schedule.
Measured from the completion announcement
Evidence-backed intervals only. Healthy reads do not prove complete collection.
What is known, and what still needs evidence.
7 eligible events, 0 with comparable scope and exact timing.
Ten events are a minimum, not a reliability guarantee. 0 records have comparable scope and exact timing.
Read status describes the current connection, not historical completeness.
24h · 0 / 0 verified complete windows contained a reset
48h · 0 / 0 verified complete windows contained a reset
| Date (UTC) | Providers and resets |
|---|---|
| 2026-10-11 | Codex: 1 |
| 2026-10-13 | Codex: 1 |
| 2026-10-17 | Codex: 1 |
| 2026-10-20 | Codex: 1 |
| 2026-10-25 | Codex: 1 |
| 2026-10-28 | Codex: 1 |
Verified provider quota resets only, not personal payments.
6 consecutive reset intervals within providers
Verified public resets only. Date-only records are excluded from interval calculations.
7 eligible events, 0 with comparable scope and exact timing.
7 eligible events, 0 with comparable scope and exact timing.
Uses smoothed historical frequencies. Incident, campaign and release modifiers remain experimental and are excluded from published estimates.
48h · Actual archives and fully observed outcomes only
No paired observations yet. Superiority to historical daily frequency is not established.
0 horizons pending; 0 mature windows lack verified coverage. Medium bands are not high/low scored. Recall includes all fully observed archived reset windows, including abstentions; it does not measure all reset events; 48h windows can overlap.
| Time | Provider | 24h | 48h | Outcome | Evaluation |
|---|
No matching records. Calculations without an issued estimate are available under All archives.
Evidence-backed completed public resets only. Grants, plans, work in progress, withdrawals and duplicates are excluded. Announcement and execution timestamps are kept separate.
At least 10 events with comparable scope and exact timing and 10 verified complete windows. Elapsed-day estimates fall back to daily frequency when support is sparse; inadequate support means no estimate.
Unknown periods are not negative outcomes. Beta(1,1) smoothing avoids zero-frequency certainty. Display bands are not confidence intervals. Experimental multipliers and invented backcasts are excluded.
These are reconciliation records, not independent events. Related records can describe one action. Historical imports send no alerts and create no past forecasts.
