Game Methods¶
If you just want the Nash Equilibria of a relatively small game, it is easiest to just call find_all_equilibria. However, if the game is too large for the call to complete in a reasonable time, or if you just feel like playing around, it might be worthwhile calling some of the other methods.
carnate_profile(self)¶
Convert a strtegy profile including one or more probability parameters to one with simple floats. When sympy finds a player is indifferent between two or more actions given the actions of the other players it will list a family of equilibria with a payoff dict like
[{0:p0, 1: 1-p0} ... ]
carnate profile essentially just sets the probability parameters to zero. Since the players are indifferent, this will not change the final payoffs.
find_all(self)¶
Finds all Nash Equilibria. Reutrns a list of lists of dicts. Each element of the outer list corresponds to a Nash Equilibium. Each element of the inner list represents the stratey profile for a particular player in that Equlibrium:
from pymnash.sample_games import stag_hunt
stag = stag_hunt(3, 2)
ne = stag.find_all_equilibria()
print(ne)
The results should be
[[{0: 1}, {0: 1}, {0: 1}], [{1: 1}, {1: 1}, {1: 1}], [{0: 0.911437827766148, 1: 0.0885621722338523}, {0: 0.911437827766148, 1: 0.0885621722338523}, {0: 0.911437827766148, 1: 0.0885621722338523}]]
This represents 3 equilibria, one where all players play action 0 (chase rabbits) one where all players play action 1 (hunt the stag), and one where the players all hunt the stag or chase rabbots with equal probabilities.
find_support_equilibria(self, support)¶
Tries to find nash equilibria with the given support. Support is a list of lists, outer list is per player, inner list is the actions that player will perform with nonzero probability.:
from pymnash.sample_games import battle_of_genders
battle = battle_of_genders(3)
ne = battle.find_support_equilibria([[0,1], [0,1], [0,1]])
The results should be
[[{0: 0.750000000000000, 1: 0.250000000000000}, {0: 0.250000000000000, 1: 0.750000000000000}, {0: 0.500000000000000, 1: 0.500000000000000}]]
find_pure(self, simple=True)¶
Finds Nash Equilibria in pure strategies. If “simple” is true, the result will be in a simplified format becuase there are no probabilities: The result is just a list of tuples, because there are no probabilities. Each element of the outer list represents an equilibrium, each element of the tuple is the actions preformed by the players in order.
If simple is False, the result will be in the same format as for find_all_equilibria:
from pymnash.sample_games import battle_of_genders
battle = battle_of_genders(3)
ne = battle.find_pure()
print(ne)
The results should be
[(0, 0, 0), (2, 1, 0), (1, 2, 0), (2, 0, 1), (1, 1, 1), (0, 2, 1), (1, 0, 2), (0, 1, 2), (2, 2, 2)]
The pure Nash Equilibria represent the three cases where all players coordinate on the same action, and also the 6 more where no two players play the same action.
get_profile_payoffs(self, profile)¶
Given a strategy profile dictof player actions, return an array showing the payoffs for all players.
is_nash(self, profile)¶
Check if the supplied strategt profile is a Nash Equilibrium. The profile is a list of dicts, each dict is the strategy profile for one player. This is the same as the output from find_all, so if you plug in a Nash Equilibrium you got from find_all, it should return True.:
from pymnash.sample_games import stag_hunt
stag = stag_hunt(3, 2)
profile = [{0: 0.911437827766148, 1: 0.0885621722338523}, {0: 0.911437827766148, 1: 0.0885621722338523}, {0: 0.911437827766148, 1: 0.0885621722338523}]
print(stag.is_nash(profile))
Should print True. The pitfalls of floating-point math may occasionally lead to this function giving the wrong answer.
iesds(self)¶
Attempts to find strictly dominated strategies. Returns a list of lists, the outer list corresponds to a player, the inner list is dominated strategies (so any Nash Equilibria do not involve those stratgeies):
from pymnash.sample_games import reducible
red = reducible(3)
print(red.iesds())
print(red.find_all_equilibria())
output:
[[1, 2], [1, 2], [1]]
[[{0: 1}, {0: 1}, {0: 1}]]
For this game, players 0 and 1 each have 3 possible actions, but actions 1 and 2 are dominated. Player 2 only has 2 actions, but action 1 is dominated. So the only Nash Equilibrium is all players playing action 0 with probability 1.
This method currenly only searches for actions which are either dominated by a singek action or by a combination of 2 acions. So for games like the mixed_dom sample game where it may take a larger combination of actions to dominate an action, this method will fail to find the dominated action.
one_player_payoffs(self, others)¶
Given strategy profiles of all the other players, find the payoffs for each strategy for the one remaining player. Others is a 3-deep list, players and strategy mix for each player:
from pymnash.sample_games import detente_of_genders
agame = detente_of_genders(4,2)
others = [[1, [[0,1]]], [2, [[0, 0.5], [1, 0.5]]], [3, [[0, 0.5], [1, 0.5]]]]
print(agame.one_player_payoffs(others)
output:
[5.0, 0.5]
In this example, player 1 is playing action 0 (probability 1) and players 2 and 3 are each playing actions 0 and 1 with probability 0.5. Output is the payoffs for player 0. If the specified others payoff is a mixed Nash Equilibrium (it is not in this case), then payoffs for the player would have to be equal for more than one action.