Validated: two blind graders, an adjudicator on every disagreement, and an agent audit of a fixed sample (D20).
| Publication | Published | Predictions | Graded | Hit rate | Status |
|---|---|---|---|---|---|
| 2025 | 2025-01-07 | 4 | complete | ||
| 2024 | 2024-01-09 | 4 | complete | ||
| 2023 | 2023 | 4 | complete | ||
| 2022 | 2022-03 | 4 | complete | ||
| 2021 | 2021 | 5 | complete | ||
| 2020 | 2020-02 | 5 | complete | ||
| 2019 | 2019-02 | 5 | complete | ||
| 2018 | 2018-02 | 5 | complete | ||
| 2017 | 2017-01 | 5 | complete | ||
| 2016 | 2016-02 | 5 | complete | ||
| 2015 | 2015-02 | 5 | complete | ||
| 2014 | 2014-01 | 6 | complete | ||
| 2013 | 2013-02 | 7 | complete | ||
| 2012 | 2012-01 | 6 | complete | ||
| 2011 | 2011-01 | 8 | complete |
Rule D33. Each trend is compared with the publisher’s next two editions: kept under the same name (persisted), kept under a new name (renamed), dropped (faded), or dropped and brought back later (recycled). This measures the publisher’s own consistency, not whether a trend came true, and is never a hit rate.
0%
persisted, 0 of 69 (0% to 5%)
38%
renamed, 26 of 69 (27% to 50%)
0%
recycled, 0 of 69 (0% to 5%)
62%
faded, 43 of 69 (51% to 73%)
Still open (fewer than two later editions): 8. Audit of the agreed rename checks: 50 checked, 2% residual error.