Sean May

Activity

Chronicle

• On 30 September 2026, Sean May drafted project planning documents and outreach materials.^[1][2]
• On 30 September 2026, he completed the engine specification and implementation plan.^[3][4][5][6]
• On 30 September 2026, he built the core engine workspace, types, board topology, and fixed action space encoding.^[7][8][9]
• On 30 September 2026, he implemented board generation, random number generation, and configuration.^[10]
• On 30 September 2026, he added the state machine, event dispatch, and setup phase.^[11]
• On 30 September 2026, he implemented rolling, production, sevens, and robber mechanics.^[12][13]
• On 30 September 2026, he added building and end-turn logic, longest road, largest army, and win checking.^[14][15]
• On 30 September 2026, he implemented development cards, maritime trades, and domestic trade protocol.^[16][17][18]
• On 30 September 2026, he added the observation model and Game wrapper to expose engine state.^[19][20]
• On 30 September 2026, he created random playouts and property-based tests to validate engine correctness.^[21][22]
• On 30 September 2026, he formatted the codebase and established benchmarks and performance baselines.^{[23][24][25][26]}
• On 30 September 2026, he exposed all public facts in Observation and added error handling for invalid chance outcomes.^[27][28]
• On 30 September 2026, he fixed longest-road accounting during setup and implemented incremental longest-road recomputation for performance.^[29][30]
• On 30 September 2026, he added golden tests for topology numbering, action IDs, and game traces.^[31][32]
• On 1 October 2026, he established head-to-head performance comparison against Catanatron and catan-rl with baseline results.^[33]
• On 1 October 2026, he added reference oracle tests for longest road and phase-level profiling.^[34]
• On 1 October 2026, he optimized longest-road, legality, and trade logic with bitmask operations and table-driven approaches.^[35]
• On 1 October 2026, he benchmarked the optimized engine and confirmed it beat catan-rl in per-step performance.^[36]
• On 1 October 2026, he created the second plan covering bots, native arena, PyO3 bindings, statistics, and CLI.^[37]
• On 1 October 2026, he implemented the Bot trait, registry, and RandomBot.^[38]
• On 1 October 2026, he added GreedyBot, native arena, PyO3 bindings, pre-roll development cards, Road Building playability, snapshots, and invariant checks.^[39]
• On 1 October 2026, he created the alphasettler CLI, arena, and stats modules; added Catanatron differential testing and oracle arena.^[40]
• On 1 October 2026, he documented plan 3 covering oracle translation, created oracle results, and updated the done-criteria checklist.^[41]
• On 1 October 2026, he wrote the ISMCTS search bot specification and sub-project 3a design document.^[42][43]
• On 1 October 2026, he enhanced the bot arena to feed each seat its redacted event log and added diagnostics.^[44]
• On 1 October 2026, he created the search crate skeleton and implemented exact belief modeling over hidden development card steals.^[45]
• On 1 October 2026, he refined belief modeling to cover joint hands exactly and updated Monopoly events to record each victim's loss.^[46]
• On 1 October 2026, he implemented trade-constrained belief sets and tested them in isolation.^[47]
• On 1 October 2026, he added sampled world generation from a validated template and enabled engine State reseeding.^[48]
• On 2 October 2026, he implemented exact on-turn VP dealing and added shared engine deck helpers.^[49]

Areas of work

References

1. ^ Commit c5dd5d7: "Initial commit: gitignore and Colonist outreach draft"
2. ^ a b Commit 91572c8: "Spec: engine architecture and internals (sections 1-2)"
3. ^ a b Commit 354a8cc: "Spec: correctness testing, benchmark harness, done criteria"
4. ^ a b Commit 41a9f50: "Spec: mark approved"
5. ^ a b Commit bb9f8ab: "Plan 1: Rust engine implementation plan"
6. ^ Commit 8ea0f4b: "Plan 1: tighten secret-scan regex to real key formats"
7. ^ a b Commit 25f1544: "engine: workspace and core types"
8. ^ a b Commit df781c1: "engine: board topology"
9. ^ Commit df44bea: "engine: fixed action space encoding"
10. ^ a b Commit de6ba4f: "engine: rng, config, and board generation"
11. ^ a b Commit 62fd546: "engine: state, events, dispatch, and setup phase"
12. ^ Commit b67d73f: "engine: rolling, production, sevens, robber"
13. ^ Commit ba8ded8: "engine: validate forced roll outcomes"
14. ^ Commit a8b6fef: "engine: building and end turn"
15. ^ Commit 3a9f513: "engine: longest road, largest army, win check"
16. ^ Commit cce3b3c: "engine: development cards"
17. ^ Commit 92b15e3: "engine: maritime trades"
18. ^ Commit d32c1ea: "engine: domestic trade protocol"
19. ^ a b Commit bac960c: "engine: observation and Game wrapper"
20. ^ a b Commit f93e49d: "engine: Game::apply normalizes only Year of Plenty order"
21. ^ a b Commit d454a54: "engine: random playouts and property tests"
22. ^ a b Commit a406b38: "engine: strengthen property tests (every action, independent VP, draw path)"
23. ^ a b Commit 255cd18: "engine: cargo fmt"
24. ^ a b Commit bd4fe70: "engine: benchmarks, throughput example, perf baseline"
25. ^ a b Commit 946414a: "engine: bench a real Main-phase position for legal_actions"
26. ^ a b Commit 79b37d0: "engine: bench a rich Main-phase position"
27. ^ a b Commit 7d6a49b: "engine: expose every public fact in Observation"
28. ^ a b Commit 8df82c3: "engine: validate forced chance before mutating; ApplyError"
29. ^ a b Commit 419cd07: "engine: count setup roads toward longest road; cache invariant"
30. ^ a b Commit 780640a: "engine: recompute longest road incrementally"
31. ^ Commit 99a55a7: "engine: golden pins for topology numbering and action ids"
32. ^ Commit cc91f97: "engine: golden trace test locking behavior for optimization work"
33. ^ a b Commit 166c5d0: "bench: head-to-head harness vs catan-rl and Catanatron, baseline results"
34. ^ a b Commit f4c1caf: "engine: longest-road reference oracle test and phase profiler example"
35. ^ Commit d140be6: "engine: bitmask road/settlement legality, table-driven offers, compact longest-road search"
36. ^ a b Commit d3447b6: "bench: post-optimization comparison (beats catan-rl per step in both trade-menu brackets)"
37. ^ a b Commit e0935f7: "plan 2: bots, native arena, PyO3 bindings, stats and CLI"
38. ^ a b Commit b255370: "bots: Bot trait, registry, RandomBot"
39. ^ a b c Commit a35e51d: "engine, bots, bindings: GreedyBot, native CRN arena, PyO3 module; pre-roll dev cards, Road Building playability, snapshots, invariants"
40. ^ a b Commit 1e88665: "alphasettler + oracle: stats, arena, CLI; Catanatron differential test and Catanatron arena"
41. ^ a b Commit 1f02201: "docs: plan 3, oracle results and done-criteria checklist, comparison fix"
42. ^ a b Commit 1012065: "spec: sub-project 3a, ISMCTS search bot"
43. ^ a b Commit be3f233: "plan 4: ISMCTS search bot (sub-project 3a); spec updated to match the code survey"
44. ^ a b c Commit 2490291: "bots: Bot::observe and diagnostics; arena feeds each seat its redacted event log"
45. ^ a b c Commit 5971a4e: "search: crate skeleton and exact belief over hidden steals"
46. ^ a b Commit 9b7975f: "plan 4, spec 3a: exact belief over joint hands; Monopoly event records each victim's loss"
47. ^ a b Commit b7d97a4: "search: trade offers and accepts constrain the belief"
48. ^ a b c Commit fa37baf: "search: sampled worlds from a validated template; engine State::reseed"
49. ^ a b c Commit b16bdea: "search: exact on-turn VP deal; shared engine deck helper"
50. ^ Commit 33785ba: "search, bots: final-review minors (legal root move, one reset per decision, failed observe keeps states, full config in records)"
51. ^ Commit e8af00c: "search: uniform deal of the cards left after the on-turn draw"
52. ^ Commit e0d9cec: "search: single-observer ISMCTS with PUCT, chance nodes and leaf batching"
53. ^ Commit b447ca5: "search: condition every opponent's hidden VP below the win"
54. ^ a b Commit 6a08cac: "search: tests for hidden chance, virtual loss, max^n backup and terminal values"
55. ^ Commit 82790e0: "bots: heuristic evaluator and IsmctsBot (ismcts, ismcts@N, ismcts@N+rD)"
56. ^ Commit d3a654d: "bots: exact belief-reset accounting in IsmctsBot"
57. ^ a b Commit 484eaaf: "selfplay: IsmctsBot self-play records, heuristic weight fit, bindings and CLI"
58. ^ a b Commit 6ec3227: "selfplay: stream records batch by batch; fit argument checks"
59. ^ a b Commit a1be6d8: "bots: report belief resets from the native arena and self-play"
60. ^ Commit 5b534c8: "oracle: translate Catanatron actions into events for bots playing inside Catanatron"
61. ^ Commit 12dfc02: "docs: 3a search speed, tuning rounds and results against Catanatron"
62. ^ Commit baf5f69: "docs: 3a results rerun on the final bot"