6
benchmark games
Casual surfaces, hidden ML structure.
Judgment is the benchmark.
The score just reveals it.
A condensed read on where ML-style validation and modeling outperform generic agent heuristics.
100%
ML win rate
Modeling beat heuristics in every game.
5
core methods
Validation, planning, graph learning, causality, backtesting.
1
benchmark signal
The decisive edge is knowing what to trust.
Benchmark summary
Score table
Final outcomes, condensed.
| Game | Metric | Generic | ML | Winner | ML advantage |
|---|---|---|---|---|---|
| Across all six games, the ML agent wins by identifying hidden structure before acting on it. | |||||
Detailed benchmark
The games and replays
Each entry shows the hidden problem, both approaches, and the replay.
Exact logs and reconstructed turn-by-turn series are preserved from the source report.