Run e861616e-8100-43e6-a8d3-39047b5b6669
Status: completed. Experiment: a5e9b3f1-8ace-418f-b8e8-bd54c825c5c9. Deterministic paired seeds, every seed played once per rotation.
Planned: 80. Attempted: 80. Completed: 80. Incomplete: 0. Unplayed: 0.
| Slot / policy | Wins | Losses | Completed | Win rate | 95% Wilson interval | Mean points |
|---|---|---|---|---|---|---|
| 0 / ntuple-leaf | 46 | 34 | 80 | 57.5% | 46.6%–67.7% | 8.28 |
| 1 / expectimax-v2-plan | 25 | 55 | 80 | 31.2% | 22.2%–42.1% | 6.90 |
| 2 / eta | 5 | 75 | 80 | 6.2% | 2.7%–13.8% | 6.10 |
| 3 / fast | 4 | 76 | 80 | 5.0% | 2.0%–12.2% | 6.03 |
Paired contrast slot 0 minus slot 1: +0.263 wins per game over 20 seeds (95% interval +0.094 to +0.431).
Manifest SHA-256: 47ce59af9a5f25a6ce6bd8bfa29dbec1b254d64b6d142edade7f6948178e49c5.
Inventory SHA-256: d1b575438b65a60ab7fa922b99084c8647f2a31d4a34c4fe6531ee7544693705.
Artifacts are local under runs/e861616e-8100-43e6-a8d3-39047b5b6669/; no remote durability is claimed until an archive receipt exists.
Intervals are descriptive per competitor slot across completed games; outcomes within a game are dependent. Incomplete games are not losses. The paired contrast is the per-seed win difference between slots 0 and 1 with a normal-approximation interval over seeds.