Scrub precomputed multi-seed results across named games. Compare algorithms, then change tell-scale
to see how side-channel strength shifts measured means — without claiming equilibria.
Explore a game
Select a game to load its teaching blurb and plots.
Algorithm
Mean
95% CI
p vs 0
Tell-scale sensitivity
Updates from cached tell-sensitivity series only — no GPU, no retraining.
Steps: 0.0 · 0.5 · 1.0 · 1.5
Plots
Loading featured plots when available…
Limitations
Reward is dense oracle classification, not a true payoff matrix.
Environments are parametric synthetics sharing one harness structure.
No adaptive opponent / self-play in the shared evaluation path.
Positive means concentrate on simple 2-action coordination games.