An open‑source Solitaire simulator lets developers brute‑force optimal play sequences. Version 1.0 offered a naïve move set (s2g, b2g, b2b, s2b) with no smart branching. Version 1.2 reordered the moves to s2g, b2b, b2g, s2b, pushing the win rate from 7.915 % to a current record of 8.590 %. Developers can clone the repo, run ant clean compile, and immediately start experiments.

Run the jar with simple flags: --one or --three to set the turn size, --attempts to define simulation count, and --seed for reproducible shuffles. A typical benchmark—one million games using --three, --attempts 1000000 and a fixed seed—finishes in under an hour on an M2 MacBook Air. Debug mode streams each game, while normal runs emit only errors and final win statistics.

The project demonstrates how exhaustive enumeration can squeeze marginal gains from a classic card game, offering a sandbox for AI researchers testing Monte‑Carlo or reinforcement‑learning strategies. By exposing move ordering and seed control, it provides a repeatable benchmark that bridges hobbyist curiosity and formal algorithmic study. The latest record of 8.590 % stands as the highest publicly documented success rate.