jidt/course/Module03-MutualInformation/ScissorsPaperRockAnalysis-M.../completed/computeConditionalEntropyFo...

50 lines
1.9 KiB
Matlab
Executable File

% function [calculatedEntropy, winRate, lossRate] = computeConditionalEntropyForPlayer(name)
%
% Compute the conditional entropy of moves for a given player, conditioned on
% their previous move across all games/iterations
%
% Input:
% - name of the player
function [calculatedEntropy, winRate, lossRate, numGames] = computeConditionalEntropyForPlayer(name)
% Step 1: load all of the player's games' data:
games = loadGamesForPlayer(name);
% Step 2: the player's moves are in the first column, pull these from
% each game into arrays of samples that we can compute conditional entropy on:
nextMoves = [];
previousMoves = [];
results = [];
for gameIndex = 1:length(games)
% Load data from game gameIndex into the variable game
game = games{gameIndex};
% First column of game is the player's move, second is opponent's
% and third is the result.
moves = game(:,1);
playersResults = game(:,3);
% Append this player's moves to the array we're storing over all iterations.
% TAKE CARE: Can we take all samples here, or only a limited number that
% we're able to match up properly to compute conditional entropy?
nextMoves = [nextMoves; moves(2:end)];
previousMoves = [previousMoves; moves(1:end-1)];
% Append this player's results to the array over all iterations:
% Which results do we want here -- those of the previous iteration or this one?
results = [results; playersResults(2:end)];
end
% Step 3: compute the condtional entropy for this player's moves using our existing scripts:
calculatedEntropy = conditionalentropyempirical(nextMoves, previousMoves);
% Step 4: compute the win and loss rates:
winRate = sum(results == 1)./length(results);
lossRate = sum(results == -1)./length(results);
numGames = length(results);
if (nargout == 0)
fprintf('Conditional entropy for %s over %d iterations: %.4f bits\n', ...
name, length(nextMoves), calculatedEntropy);
end
end