BUTT DESTROYER 9001 v1.1

This program has been disqualified.


AuthorJFreegman
Submission date2012-07-26 08:34:14.089289
Rating6924
Matches played44
Win rate72.73

Source code:

# Author: JFreegman
# Contact: JFreegman@gmail.com
# Date: July 25, 2012
# v1.1
# Update from v1.0: pattern recognition improved and optimized

# This is the BUTT DESTROYER 9001 v1.1, a rock paper scissors playing bot 
# originally written for Udacity's CS212 rock, paper scissors tournament and 
# adapted for RPSContest.com.

# All code is written from scratch. The general idea is based off Iocaine Powder 
# by Dan Egnor (http://ofb.net/~egnor/iocaine.html).

# What this bot attempts to do is create a prediction heuristic based on
# frequency of the opponent's moves and patterns in their history. The bot will
# keep track of the performance of a number of strategies and meta-strategies 
# while playing the move that its current most successfull strategy comes up with.

import random

def get_history_match():
    """
    Searches moves history and tries to find previous patterns that match
    the last n moves (n starting at half the number of moves and decrementing).
    If match found, returns the move made after the last pattern match.
    """
    start = len(PATTERNS) - min(len(PATTERNS) / 2, 200)
    end = len(PATTERNS)
    for i in xrange(start, end):
        partition = PATTERNS[i:end]
        match = PATTERNS[:-1].find(partition)
        if match != -1:
            return PATTERNS[match+len(partition)]
    return random_weapon()

def get_probs(total_moves, n):
    """
    Returns a dictionary containing the probabilities that the opponent
    will make a given move based on their last n moves.
    """
    last = get_move_freq(total_moves[-n:])
    probs = {}
    probs['R'] = float(last['R']) / last['total'] * 100
    probs['S'] = float(last['S']) / last['total'] * 100
    probs['P'] = float(last['P']) / last['total'] * 100
    return probs

def get_move_freq(moves):
    """
    Returns a dictionary containing frequencies of moves, as well as
    a count for the total number of moves
    """
    mov_freq = {'R': 0, 'P': 0, 'S': 0}
    count = 0
    for move in moves:
        if move == 'R':
            mov_freq['R'] += 1
        elif move == 'P':
            mov_freq['P'] += 1
        elif move == 'S':
            mov_freq['S'] += 1
        else:
            raise ValueError, 'Invalid move'
        count += 1
    mov_freq['total'] = count
    return mov_freq

def winning_move(m):
    "Returns the move that beats m"
    d = {'R': 'P', 'P': 'S', 'S': 'R'}
    return d[m]

def losing_move(m):
    "Returns the move that loses to m"
    d = {'R': 'S', 'P': 'R', 'S': 'P'}
    return d[m]

def random_weapon():
    "Randomly chooses R, P or S"
    moves = ['R', 'P', 'S']
    return random.choice(moves)

if not input:
    # Keeps history of each strategy's move for each round
    STRAT_HISTORY = [{}]


    # Keeps track of the success rate of each strategy
    STRAT_SUCCESS = {'freq20': 0, 'c1_freq20': 0, 'c2_freq20': 0, 'hist': 0, 'random': 0,
                     'c1_hist': 0, 'c2_hist': 0, 'c1_freq100': 0, 'c2_freq100': 0,
                     'freq100': 0, 'freqtot': 0, 'c1_freqtot': 0, 'c2_freqtot': 0}

    # Keeps track of opponent move patterns
    PATTERNS = ""
    opp_moves = []
    output = random_weapon()
    my_moves = [output]
else:
    # update strategy success rates based on last round results
    opp_moves.append(input)
    last_opp_move = input
    beat_opp = winning_move(last_opp_move)
    lose_opp = losing_move(last_opp_move)
    last_strats = STRAT_HISTORY[-1]
    for s in last_strats:
        if last_strats[s] == beat_opp:
            STRAT_SUCCESS[s] += 1
        elif last_strats[s] == lose_opp:
            STRAT_SUCCESS[s] -= 1

    PATTERNS += last_opp_move

    # get opponent's most probable move based on frequency (last 20, 100 and
    # full history) and history pattern matches
    opp_freq20 = get_probs(opp_moves, 20)
    opp_prob_f_20 = max(opp_freq20, key=opp_freq20.get)
    opp_freq100 = get_probs(opp_moves, 100)
    opp_prob_f_100 = max(opp_freq100, key=opp_freq100.get)
    opp_freqtot = get_probs(opp_moves, len(opp_moves))
    opp_prob_f_tot = max(opp_freqtot, key=opp_freqtot.get)
    opp_prob_h = get_history_match()

    # naive moves for each strategy
    my_move_freq20 = winning_move(opp_prob_f_20)
    my_move_freq100 = winning_move(opp_prob_f_100)
    my_move_freqtot = winning_move(opp_prob_f_tot)
    my_move_hist = winning_move(opp_prob_h)
    random_move = random_weapon()

    # counter-moves in case opp predicts my naive moves
    c1_move_freq20 = losing_move(my_move_freq20)
    c1_move_freq100 = losing_move(my_move_freq100)
    c1_move_freqtot = losing_move(my_move_freqtot)
    c1_move_hist = losing_move(my_move_hist)

    # counter-counter moves in case opp predicts my first counter
    c2_move_freq20 = losing_move(c1_move_freq20)
    c2_move_freq100 = losing_move(c1_move_freq100)
    c2_move_freqtot = losing_move(c1_move_freqtot)    
    c2_move_hist = losing_move(c1_move_hist)

    # dict of all available strategies and their move
    strats = {'freq20': my_move_freq20, 'freq100': my_move_freq100, 
              'hist': my_move_hist, 'random': random_move,
              'c1_freq20': c1_move_freq20, 'c2_freq20': c2_move_freq20, 
              'c1_freq100': c1_move_freq100, 'c2_freq100': c2_move_freq100, 
              'c1_hist': c1_move_hist, 'c2_hist': c2_move_hist,
              'freqtot': my_move_freqtot, 'c1_freqtot': c1_move_freqtot,
              'c2_freqtot': c2_move_freqtot}

    # Pick the strategy with the highest current success rate
    strat = max(strats, key=lambda x: STRAT_SUCCESS[x])
    output = strats[strat]
    my_moves.append(output)
    STRAT_HISTORY.append(strats)