Records
Structured prior against uniform setups, greedy in both seats
Registration 5d4a4b4f-3d46-4d75-8354-2e2492377063 (study setups/structured-prior), 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 f64c784, research commit 79af9a4.
Command:
cd server && cargo run --release -p armies-arena --bin matchup -- --red 'greedy-structured=greedy:{"setup":"structured"}' --blue 'greedy-uniform=greedy:{"setup":"uniform"}' --games 400 --seed 0 --rules training --threads 8 --moves --out ../research/runs/setup-prior/greedy.json --record-dir ../research/runs/setup-prior/records-greedy --tag 5d4a4b4f-3d46-4d75-8354-2e2492377063
cd research && python3 scripts/paired_contrast.py runs/setup-prior/greedy.json| Player | Wins | Draws | Losses | Effective | 95% Wilson | Mean plies | ms/move |
|---|---|---|---|---|---|---|---|
| greedy-structured | 260 | 52 | 88 | 0.715 | 0.669 to 0.757 | 202 | 0.002 |
| greedy-uniform | 88 | 52 | 260 | 0.285 | 0.243 to 0.331 | 202 | 0.002 |
| ending: flag_captured | 334 | ||||||
| ending: no_moves | 14 | ||||||
| ending: quiet_limit | 52 |
Per-seed paired contrast for the structured side (score as Red plus score as Blue minus one): mean 0.430, 95% normal interval 0.339 to 0.521 over 200 seeds (121 positive, 48 split, 31 negative). The structured side won 134 games as Red and 126 as Blue. Draw share 13.0%. Wall time 0.1 s on 8 threads; this run shared the machine with other agents, which does not change deterministic scores. An identical-outcome run alone measured the same 0.002 ms per move. Sixteen of the 400 games (4.0%) ended within 20 plies; the uniform side lost all sixteen.
Provenance
Archived JSON: runs/setup-prior/greedy.json (SHA-256 0606d697c4fc1652ea49c64fdc266fd5bc62c3a88762feaf555bd13483007763), 18.3 MB with every game's move list and per-ply detail, so it is not copied next to this record. Games planned 400, attempted 400, completed 400, interrupted 0. The curated replay library holds seed 0 from both seats (2 records under records/games/structured-v-uniform/). No deviation from the registration; the budget of 2,000 ms per move was met with a mean of 0.002 ms.