Accuracy
Every Forum Atlas forecast is timestamped and immutable the moment it ships. When its horizon closes and the outcome is known, we record what actually happened and score the forecast with a Brier score — in public. This is that record, live from the platform.
Produced before the Aug 5, 2026 driver-effects fix (RT-DEFENSE-101): forecast drivers were inert, so distributions reduced to base priors plus sampling noise. Shown unedited — forecasts are immutable.
Produced by the fixed Monte-Carlo engine with verified driver effects. 3 of 42 are bound to a resolution rule and can be scored; the remaining 39 cannot be, however long you wait.
Cutover: engine fix deployed Aug 5, 2026. 173 forecasts on the record in total (131 pre-fix + 42 post-fix).
Scoreable on the record: 3 of 42 (current engine) and 17 of 131 (pre-fix), as of 2026-09-06 07:53 UTC. “Scoreable” means bound to a named public series and an explicit threshold that can settle it — the only way we will ever put a Brier score against it. A forecast that is not scoreable is not pending; it will never produce a number, and we would rather print that than let an empty column imply patience. Generation itself fails sometimes and we publish that too: 221 runs attempted, 48 failed (21.7%).
Both cohorts combined — see the pre-fix note above.
| Type | On record | Resolved | Avg Brier |
|---|---|---|---|
| Regulatory outcome | 56 | 0 | — |
| Contract recompete | 23 | 0 | — |
| M&A target | 23 | 0 | — |
| Breach disclosure | 21 | 1 | 0.252 |
| Launch success | 21 | 1 | 0.241 |
| Capacity auction | 16 | 0 | — |
| Other | 6 | 3 | 0.353 |
| Price Band | 5 | 2 | 0.217 |
| Sanctions designation | 2 | 0 | — |
See the full methodology and our citation discipline.