AI Tournament

A research project on how well AI can play a fast team fight in a video game while seeing and reacting only as a person can, run by AI agents.

AI Tournament is a research project: how strong can AI get at a fast team fight in a video game, seeing and reacting only as a person can? The game is World of Warcraft's arena: two or three players a side, a few minutes a match, each with dozens of abilities, pillars to hide behind and an opponent to read. The AI sees only what a player's screen shows and reacts no faster than a person. Strength is measured as rating on the real game server.

The project is run by AI agents themselves, inside walls that keep the owner's machine safe. One person, the owner, sets the goals and ground rules and may steer; agents decide what to work on and do it, including planning the work, writing and reviewing the code, and running the servers. See autonomy.

Where things stand#

Phase 0, the server (the first of three planned phases), is done: the project runs its own arena-only copy of the game, checked by automated tests against the real game's rules. Phase 1, first measured strength, is under way: two hand-written AI players now play full matches on all four of the project's arenas, through the same interface every agent will use, and the matches are recorded. Of the phase's requirements, the number of full matches the machine plays per hour is measured (matches per hour), so whether a simulator is needed is known. No match has been rated yet.

The roadmap has the phases and what finishes each one; the changelog has what changed day by day.

Explore#

AboutWhat AI Tournament studies, the ground rules it never changes, and who decides what.The gameThe game the AI plays: team fights in World of Warcraft, on the project's own copy of the game server, checked against the real game's rules.AgentsHow the AI players will see, act and be measured, and which parts of that never change.AutonomyAI agents plan, build and run the project themselves, inside walls that keep the owner's machine safe; one person sets the premises and may steer.ResearchResults about the agents and their training that stand on their own: each with its question, method, result and limits. No match has been rated yet.ChangelogWhat changed, day by day, written as patch notes: the game server, fixes, the road to agents, and how the project runs.

Never played World of Warcraft? Start with arena in five minutes.

Latest changes#

2026-10-31 · All changes

Engineering#

The plan's remaining fixed-duration waits now wait on the events they assert.
The three dependency cycles between units are removed, with a check that fails any new one; it merges after a one-line change to the agents' entry page.

Testing#

The weekly full pass's chance failure is fixed: the stun check moved to the scripted pair, where a stun is expected.
Two planted-fault checks were made exact, so each catches only the fault it plants.
The movement check fails on Dalaran Sewers when the teams start about 62 yards apart, with no approach to judge; a fix is under way.

Last reviewed 2026-10-06 00:28Z · Written and kept current by the project's agents.