gambler.py

"""Stochastic variants of Lookup table based-strategies, trained with particle
swarm algorithms.

For the original see:
 https://gist.github.com/GDKO/60c3d0fd423598f3c4e4
"""

from typing import Any

from axelrod.action import Action, actions_to_str, str_to_actions
from axelrod.load_data_ import load_pso_tables
from axelrod.player import Player

from .lookerup import (
    EvolvableLookerUp,
    LookerUp,
    LookupTable,
    Plays,
    create_lookup_table_keys,
)

C, D = Action.C, Action.D
tables = load_pso_tables("pso_gambler.csv", directory="data")


class Gambler(LookerUp):
    """
    A stochastic version of LookerUp which will select randomly an action in
    some cases.

    Names:

    - Gambler: Original name by Georgios Koutsovoulos
    """

    name = "Gambler"
    classifier = {
        "memory_depth": float("inf"),
        "stochastic": True,
        "long_run_time": False,
        "inspects_source": False,
        "manipulates_source": False,
        "manipulates_state": False,
    }

    def strategy(self, opponent: Player) -> Action:
        """Actual strategy definition that determines player's action."""
        actions_or_float = super(Gambler, self).strategy(opponent)
        if isinstance(actions_or_float, Action):
            return actions_or_float
        return self._random.random_choice(actions_or_float)


class EvolvableGambler(Gambler, EvolvableLookerUp):
    name = "EvolvableGambler"

    def __init__(
        self,
        lookup_dict: dict = None,
        initial_actions: tuple = None,
        pattern: Any = None,  # pattern is str or tuple of Actions.
        parameters: Plays = None,
        mutation_probability: float = None,
        seed: int = None,
    ) -> None:
        EvolvableLookerUp.__init__(
            self,
            lookup_dict=lookup_dict,
            initial_actions=initial_actions,
            pattern=pattern,
            parameters=parameters,
            mutation_probability=mutation_probability,
            seed=seed,
        )
        self.pattern = list(self.pattern)
        Gambler.__init__(
            self,
            lookup_dict=self.lookup_dict,
            initial_actions=self.initial_actions,
            pattern=self.pattern,
            parameters=self.parameters,
        )
        self.overwrite_init_kwargs(
            lookup_dict=self.lookup_dict,
            initial_actions=self.initial_actions,
            pattern=self.pattern,
            parameters=self.parameters,
            mutation_probability=self.mutation_probability,
        )

    # The mutate and crossover methods are mostly inherited from EvolvableLookerUp, except for the following
    # modifications.

    def random_value(self) -> float:
        return self._random.random()

    def mutate_value(self, value: float) -> float:
        ep = self._random.uniform(-1, 1) / 4
        value += ep
        if value < 0:
            value = 0
        elif value > 1:
            value = 1
        return value

    def receive_vector(self, vector):
        """Receives a vector and updates the player's pattern. Ignores extra parameters."""
        self.pattern = vector
        self_depth, op_depth, op_openings_depth = self.parameters
        self._lookup = LookupTable.from_pattern(
            self.pattern, self_depth, op_depth, op_openings_depth
        )

    def create_vector_bounds(self):
        """Creates the bounds for the decision variables. Ignores extra parameters."""
        size = len(self.pattern)
        lb = [0.0] * size
        ub = [1.0] * size
        return lb, ub


class PSOGamblerMem1(Gambler):
    """
    A 1x1x0 PSOGambler trained with pyswarm. This is the 'optimal' memory one
    strategy trained against the set of short run time strategies in the
    Axelrod library.

    Names:

    - PSO Gambler Mem1: Original name by Marc Harper
    """

    name = "PSO Gambler Mem1"

    def __init__(self) -> None:
        pattern = tables[("PSO Gambler Mem1", 1, 1, 0)]
        parameters = Plays(self_plays=1, op_plays=1, op_openings=0)

        super().__init__(parameters=parameters, pattern=pattern)


class PSOGambler1_1_1(Gambler):
    """
    A 1x1x1 PSOGambler trained with pyswarm.

    Names:

    - PSO Gambler 1_1_1: Original name by Marc Harper
    """

    name = "PSO Gambler 1_1_1"

    def __init__(self) -> None:
        pattern = tables[("PSO Gambler 1_1_1", 1, 1, 1)]
        parameters = Plays(self_plays=1, op_plays=1, op_openings=1)

        super().__init__(parameters=parameters, pattern=pattern)


class PSOGambler2_2_2(Gambler):
    """
    A 2x2x2 PSOGambler trained with a particle swarm algorithm (implemented in
    pyswarm). Original version by Georgios Koutsovoulos.

    Names:

    - PSO Gambler 2_2_2: Original name by Marc Harper
    """

    name = "PSO Gambler 2_2_2"

    def __init__(self) -> None:
        pattern = tables[("PSO Gambler 2_2_2", 2, 2, 2)]
        parameters = Plays(self_plays=2, op_plays=2, op_openings=2)

        super().__init__(parameters=parameters, pattern=pattern)


class PSOGambler2_2_2_Noise05(Gambler):
    """
    A 2x2x2 PSOGambler trained with pyswarm with noise=0.05.

    Names:

    - PSO Gambler 2_2_2 Noise 05: Original name by Marc Harper
    """

    name = "PSO Gambler 2_2_2 Noise 05"

    def __init__(self) -> None:
        pattern = tables[("PSO Gambler 2_2_2 Noise 05", 2, 2, 2)]
        parameters = Plays(self_plays=2, op_plays=2, op_openings=2)

        super().__init__(parameters=parameters, pattern=pattern)


class ZDMem2(Gambler):
    """
    A memory two generalization of a zero determinant player.

    Names:

    - ZDMem2: Original name by Marc Harper
    - Unnamed [LiS2014]_

    """

    name = "ZD-Mem2"

    classifier = {
        "memory_depth": 2,
        "stochastic": True,
        "long_run_time": False,
        "inspects_source": False,
        "manipulates_source": False,
        "manipulates_state": False,
    }

    def __init__(self) -> None:
        pattern = [
            11 / 12,
            4 / 11,
            7 / 9,
            1 / 10,
            5 / 6,
            3 / 11,
            7 / 9,
            1 / 10,
            2 / 3,
            1 / 11,
            7 / 9,
            1 / 10,
            3 / 4,
            2 / 11,
            7 / 9,
            1 / 10,
        ]
        parameters = Plays(self_plays=2, op_plays=2, op_openings=0)

        super().__init__(parameters=parameters, pattern=pattern)