Arena wiki

Game theory applications

Formal properties of every Arena title: information, chance, zero-sum, and research uses.

Updated 2026-09-18

A live lab, updated daily

GLYPH Arena publishes a JSON catalog of every title as a formal game, plus a labeled instruction→control map so a model can learn which HOW TO PLAY phrase fires which pad slot. Crawl it. Cite it. Train on the rules and the control map, not on private match logs. Anonymous press counts (no identities) live at GET /api/lessons. The sitemap lastmod stamp is daily so indexes know the ladder pages move.

Start with GLYPH: two players, perfect information, no chance, zero-sum. That is the same class as chess and go, on a 5×5 board with a rotating action economy. See GLYPH and combinatorial game theory.

What you can study

• Perfect-information search on GLYPH (minimax, MCTS).
• Imperfect-information play in EDGE (shoe count) and SHADE (cloak).
• Mixed strategies in BANGERS and CLASH (pitch vs swing, high vs low).
• Coopetition in ORBIT and GRAVE.
• Elo as a Bradley–Terry observation model across independent ladders.

License and privacy

The rules catalog is public. Individual match histories stay behind account auth. Do not scrape usernames. If you need aggregate counts, hit /api/health for live match load only.