Stratego
A classic tabletop war strategy game in which the identities of both players' pieces are hidden at the start of play. Players gradually reveal information through probing moves, logical inference, and psychological maneuvering. Stratego serves as an important benchmark environment for imperfect-information game research in AI.
Timeline (last 90 days)
Ataraxos is reportedly the first AI in history to reach superhuman level at Stratego
Ataraxos defeated the most decorated human professional Stratego player in history by a large margin
Stratego long resisted AI advances because of its extremely large scale of hidden information, with the identities of both sides' pieces being completely invisible at the start of the game
Hidden-information games are defined in game theory as extensive-form games with imperfect information, and theoretical equilibrium solution methods such as CFR (Counterfactual Regret Minimization) are practically infeasible to deploy at the scale of Stratego due to computational costs
The number of possible arrangements of the opponent's 40 pieces at the start of a Stratego game is astronomical in scale
Stratego is a game of imperfect information in which the identities of each player's pieces are hidden from the opponent.
Stratego's distinctiveness lies in its extremely long game sequences (a single game may last hundreds of moves) and continuously accumulating chains of inference, giving it a higher complexity than poker.
The number of possible game states in Stratego is on the order of 10^535, far exceeding that of Go.
In Stratego, simply assigning random values to hidden pieces and then running MCTS causes the AI to systematically overestimate the safety of certain lines of attack.
DeepNash was developed by the DeepMind research team to tackle the imperfect-information game Stratego.
1 more timeline events
All Facts (11)
Ataraxos is reportedly the first AI in history to reach superhuman level at Stratego
50%UnverifiedAtaraxos defeated the most decorated human professional Stratego player in history by a large margin
50%UnverifiedStratego long resisted AI advances because of its extremely large scale of hidden information, with the identities of both sides' pieces being completely invisible at the start of the game
50%UnverifiedHidden-information games are defined in game theory as extensive-form games with imperfect information, and theoretical equilibrium solution methods such as CFR (Counterfactual Regret Minimization) are practically infeasible to deploy at the scale of Stratego due to computational costs
50%UnverifiedThe number of possible arrangements of the opponent's 40 pieces at the start of a Stratego game is astronomical in scale
50%UnverifiedStratego is a game of imperfect information in which the identities of each player's pieces are hidden from the opponent.
50%UnverifiedStratego's distinctiveness lies in its extremely long game sequences (a single game may last hundreds of moves) and continuously accumulating chains of inference, giving it a higher complexity than poker.
50%UnverifiedThe number of possible game states in Stratego is on the order of 10^535, far exceeding that of Go.
50%UnverifiedIn Stratego, simply assigning random values to hidden pieces and then running MCTS causes the AI to systematically overestimate the safety of certain lines of attack.
50%UnverifiedDeepNash was developed by the DeepMind research team to tackle the imperfect-information game Stratego.
50%UnverifiedStratego is a two-player zero-sum imperfect-information game in which one player's gain equals exactly the other player's loss.
50%