Records

The selected pool against uniform setups

Registration 6fb87c90-d869-46d2-8a57-eb2b367a74a0 (study setups/selection-by-self-play), run 2026-09-19 on AMD Ryzen AI Max (Strix Halo), 32 threads, 125 GB unified memory, Radeon 8060S via ROCm, Linux 6.18. Server commit 0eb994c, research commit 19c62a6. The run shared the machine with other agents' jobs, which inflates the per-move clocks but not the deterministic scores.

Command:

cd server && target/release/matchup -- --red 'pool=expectimax:{"samples":16,"depth":2,"setup":"pool:../research/content/assets/setup-pool.json"}' --blue 'uniform=expectimax:{"samples":16,"depth":2,"setup":"uniform"}' --games 400 --seed 0 --rules training --threads 8 --out ../research/runs/setup-selection/cell-pool-v-uniform.json --record-dir ../research/runs/setup-selection/records-pool-v-uniform --tag 6fb87c90-d869-46d2-8a57-eb2b367a74a0

Both seats play expectimax over sixteen sampled worlds at depth 2; the red side draws its setup uniformly from the selected 32-entry pool, the blue side uniformly over all arrangements.

Player	Wins	Draws	Losses	Effective	95% Wilson	Mean plies	ms/move
pool	357	3	40	0.896	0.863 to 0.923	178	6.037
uniform	40	3	357	0.104	0.078 to 0.138	178	4.548
ending: flag_captured	376
ending: no_moves	21
ending: quiet_limit	3

Per-seed paired contrast for the pool side: mean 0.792, 95% normal interval 0.728 to 0.857 over 200 seeds (164 positive, 31 split, 5 negative). The pool side won 176 games as Red and 181 as Blue. Draw share 0.8%. Thirty-eight games (9.5%) ended within 20 plies and every one of them was lost by the uniform side. Wall time 48.4 s on 8 threads; the budget of 2,000 ms per move was met with a mean of 6.0 ms.

The registered comparison is descriptive: on the same seeds and the same move policy the structured prior scored a contrast of 0.670 (0.589 to 0.751) against uniform setups, and the pool scored 0.792 (0.728 to 0.857), a higher point estimate with overlapping intervals.

Provenance

Archived JSON: runs/setup-selection/cell-pool-v-uniform.json (SHA-256 a24397e319d2e0312e132113256b1dc63b6cadb44bb35c735ddcda664ebe0987); per-game records under runs/setup-selection/records-pool-v-uniform/. Games planned 400, attempted 400, completed 400, interrupted 0. No deviation from the registration; no promotion rides on this cell alone.