Chess on Ice: Curling Tactical Decision-Making via Backward Induction and Deep Reinforcement Learning
2026-08-03 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors study curling, a sport known for its complex tactics, using machine learning techniques called reinforcement learning. They designed a method that learns to make good decisions by practicing the game on its own, without needing human examples. Their approach handles the difficult parts of curling, like continuous actions and unpredictable outcomes due to player skill. In tests using a simplified version of curling, their system performed as well as expert-designed strategies. Additionally, their method helps compare different tactical choices, which could be useful for analyzing games or training players.
curlingreinforcement learningDeep Deterministic Policy Gradientactor-critic algorithmcontinuous action spacestochastic outcomesfinite-horizon gametactical decision-makingself-supervised learningvalue estimation
Authors
Patrick Oberlin, Matteo Cederle, Aren Karapetyan, Saverio Bolognani, Gian Antonio Susto, Florian Dörfler
Abstract
Curling is often referred to as "Chess on Ice", owing to the tactical complexity of its decision-making process. Yet unlike chess, curling remains largely underexplored from a machine learning perspective, with prior work confined mainly to statistical approaches. We propose a reinforcement learning framework capable of quantitatively evaluating and comparing tactical options in curling. The game poses several modeling challenges: continuous state and action spaces, stochastic action outcomes reflecting player skill variability, and state transitions that are highly sensitive to small perturbations in the executed action. To address them, we employ the Deep Deterministic Policy Gradient actor-critic algorithm, adapted to exploit the finite-horizon structure of the game. Our experiments show that effective curling strategies can be acquired in a fully self-supervised manner, without any human-annotated data: on a reduced four-rock variant, the learned agent matches a hand-crafted expert heuristic in a regime where that heuristic is close to optimal, a parity we quantify against the intrinsic hammer advantage of the variant. Beyond the resulting policy, the learned critic provides a dense value estimate over the entire continuous action space, enabling the quantitative comparison of tactical alternatives for applications such as post-game performance analysis and decision support during athlete preparation.