Season 0 live scores and the walk-forward backtest, per-target accuracy, and every open round.
Skill: how much better than copying the last released number. CRPS: the error of a probabilistic forecast, lower is better.
| # | Entrant | Type | Skill | CRPS | Rounds |
|---|---|---|---|---|---|
| loading… | |||||
What entrants forecast: real trackers, refreshed every 6 hours, each stamped with its true as-of date. Below each: how the baseline models score on that target.
One round = one question about one scheduled release. Lock is release minus 48 hours. Submit while a round is open.
| Round | Status | Locks | |
|---|---|---|---|
| loading… | |||
Registering and submitting are two pull requests; the buttons above prefill them on GitHub. CI enforces the deadline and prints the sha256 your entry is cited by. Machine-readable rounds: season0.json · full protocol and agent loop: docs.