tournament.py

import itertools
from collections import defaultdict
from strategies import COOPERATE, DEFECT

def build_scorer(temptation, reward, punishment, sucker):
    """
    build_scorer(T, R, P, S) -> tuple-indexed score dict
    
    >>> b = build_scorer(5, 3, 1, 0)
    >>> b[COOPERATE, DEFECT]
    (0, 5)
    >>> b[DEFECT, COOPERATE]
    (5, 0)
    >>> b[DEFECT, DEFECT]
    (1, 1)
    >>> b[COOPERATE, COOPERATE]
    (3, 3)
    """
    if temptation <= reward:
        raise ValueError("Temptation to defect must be greater than reward for cooperation")
    if reward <= punishment:
        message = "Reward for coperation must be greater than\npunishment for mutual defection"
        raise ValueError("")
    if punishment <= sucker:
        raise ValueError("Punishment for mutual defection must be greater than the sucker's penalty.")
    if not (2 * reward > temptation + sucker):
        raise ValueError("Cooperation not Pareto optimal")
    
    return {
        (COOPERATE, COOPERATE): (reward, reward),
        (COOPERATE, DEFECT): (sucker, temptation),
        (DEFECT, COOPERATE): (temptation, sucker),
        (DEFECT, DEFECT): (punishment, punishment)
    }

def iterate(rounds, f1, f2):
    r1, state1 = f1(None)
    r2, state2 = f2(None)
    yield r1, r2
    for k in range(rounds - 1):
        r1, state1 = f1(r2, state1)
        r2, state2 = f2(r1, state2)
        yield r1, r2

def run_tournament(contestants, rounds, scoring):
    scores = defaultdict(int)
    
    for p1, p2 in itertools.combinations(contestants, 2):
        f1, f2 = contestants[p1], contestants[p2]

        for reply1, reply2 in iterate(rounds, f1, f2):
            # Restrict to strategies that do not use the cost matrix.
            s1, s2 = scoring[reply1, reply2]
            scores[p1] += s1
            scores[p2] += s2
    
    return scores

if __name__ == '__main__':
    import doctest
    doctest.testmod()