System and method for conducting a game including a computer-controlled player

Abstract

A system and method for conducting a game between at least one live player and at least one computer-controlled player includes executing a training program between at least two agents to generate probability weights correlating actions or meta-actions representing a set or sequenced set of actions with a probability that the action or meta-action will produce a game outcome meeting a specified criterion or specified criteria. A game is conducted in which at least one live player plays against at least one computer-controlled player in which the computer-controlled player selects actions at one or more of the decision nodes in the game based, at least in part, on the probability weights.

1 . (canceled) 
     
     
         2 . A method for use by a computer system for conducting a game between at least one computer-controlled player and at least one live player comprising:
 identifying, by the system, that a decision node in the game has been reached;   providing, by the system, a game state and a plurality of available actions to at least one neural network;   receiving, by the system, from the neural network, a likelihood that each of the available actions will satisfy a predetermined criterion based on the game state;   randomly selecting, by the system, one of the available actions for the computer-controlled player, wherein each of the available actions has a likelihood of being selected that is equal to the likelihood of that available action satisfying the predetermined criterion;   taking, by the system, the selected available action with the computer-controlled player.   
     
     
         3 . The method of  claim 2  further comprising:
 identifying, by the system, one of a plurality of game states at the decision node; and 
 using, by the system, a predefined probability distribution of probability weights corresponding to the identified game state for the available actions. 
 
     
     
         4 . The method of  claim 2  wherein the predefined criterion is a minimization of maximum loss criterion. 
     
     
         5 . The method of  claim 2  wherein the likelihood of being selected for each action is based on a probability weight generated by a neural network training program. 
     
     
         6 . The method of  claim 2  wherein the game is poker and wherein the available actions include fold, call, and raise. 
     
     
         7 . The method of  claim 6  wherein wagers are made at decision nodes during the game, and wherein the wagers are contributed to a pot. 
     
     
         8 . The method of  claim 7  wherein the plurality of inputs further includes a size of a pot. 
     
     
         9 . The method of  claim 7  further comprising determining the game outcome and distributing at least a portion of the pot is distributed to the live player if the live player wins the poker game. 
     
     
         10 . The method of  claim 2  wherein at least one of the available actions is a meta-action that represents a sequence of two or more actions. 
     
     
         11 . The method of  claim 10  wherein only a first portion of the sequence is executed at the decision node. 
     
     
         12 . The method of  claim 11  wherein a second portion of the sequence is executed at another decision node. 
     
     
         13 . The method of  claim 12  wherein execution of the second portion of the sequence at another decision node is conditional. 
     
     
         14 . A system for conducting a game between at least one computer-controlled player and at least one live player comprising:
 a data processor;   a data storage; and   a plurality of instructions stored in the data storage and executable by the data processor, the instructions including;   instructions for identifying that a decision node in the game has been reached;   instructions for providing a game state and a plurality of available actions to at least one neural network;   instructions for receiving, by the system, from the neural network, a likelihood that each of the available actions will satisfy a predetermined criterion based on the game state; and   instructions for randomly selecting one of the available actions for the computer-controlled player, wherein each of the available actions has a likelihood of being selected that is equal to the likelihood of that available action satisfying the predetermined criterion.   
     
     
         15 . The system of  claim 14  wherein the game is poker, and wherein the available actions include fold, call, and raise. 
     
     
         16 . The system of  claim 15  further comprising instructions for receiving wagers at decision nodes during the game, and instructions for adding the wagers are to a pot. 
     
     
         17 . The system of  claim 16  further comprising instructions for determining a game outcome and instructions for distributing at least a portion of the pot is distributed to the live player if the live player wins the poker game. 
     
     
         18 . The system of  claim 14  further comprising:
 instructions for identifying one of a plurality of game states at the decision node; and 
 instructions for using a predefined probability distribution of probability weights corresponding to the identified game state for the available actions. 
 
     
     
         19 . The system of  claim 18  wherein at least one of the available actions is a meta-action that represents a sequence of two or more actions. 
     
     
         20 . The system of  claim 19  wherein only a first portion of the sequence is executed at the decision node. 
     
     
         21 . The system of  claim 20  wherein a second portion of the sequence is executed at another decision node.