Structured multi-agent world model improves ten-hero MOBA gaming policy
Learning in Dreams, Winning in Reality: A Continuous Dyna Loop for a Ten-Hero MOBA
Artificial Intelligence
Summary
Playing a complex strategy game with many characters is hard for computers because the game is very detailed and long. The authors created a detailed simulation (a world model) that predicts how the game works and trained a strategy inside it. They found that combining imagined training with real-game testing and updates led to a strategy that wins over 70% of the time in the real game, while training only inside the simulation performed poorly. Their approach highlights how important it is to fix errors in the simulation and that the learned tactics focus on broad map control rather than close fighting.
What this means in practice
- •For game ai developers: Train competitive strategies for complex multiplayer games using continuous looping between simulation and real-game evaluation.
- •For robotic control engineers: Improve control policies by integrating learned environment models with real-world feedback to correct simulation errors continuously.
Authors
Jordy Kieto
Abstract
World models are usually judged from the inside: by prediction loss, by the return a policy earns in imagination, or by how convincing their frames look. We judge one from the outside. We learn a structured, multi-agent world model of a complete ten-hero MOBA (206 units, every hero acting every tick, games of up to 6,000 ticks), train a policy only inside it with 1,400-tick free-running imagined episodes, and measure that policy in the real game against the opponent the game ships with. The real game never provides a gradient; it provides the policy's own games as training data for the world model, and an online evaluation that selects and anchors the policy. Run as a continuous asynchronous Dyna loop, the policy wins 70.2% of real games as radiant (421 of 600; 95% CI 66.4-73.7) on seeds never used for any decision, up from 0% for dream training alone and 33.7% before the loop. It wins none as dire, and neither does the shipped opponent when it plays itself. Four findings explain the result. Model exploitation is invisible from inside the dream: every unanchored run collapsed within a few updates while no in-dream metric tracked the collapse. A world model that is accurate on its training corpus is badly wrong on the policy's own games, and Dyna repairs it there, which is worth +9.2 points of real win rate with the policy recipe held fixed. Finally, the policy inherits its world model's fidelity profile mechanic by mechanic: the model represents the macro game but not crowd control, cast timing or lethality, and the policy wins by map-wide pressure with almost no coordinated fighting. We release the world model, the dream-PPO harness, a world-model debugger, the evaluation protocol, and every policy and log.