Salvador Badillo-Rios and Verenice Mojica Goal The goal of this research project was to provide an extended analysis of 2-D Chomp using computational and mathematical means in order to provide a pattern that may aid in finding the winning strategy for all board sizes. Game Description Geometric 2-D Chomp: Two-player game Players take turns choosing a square box from an m x n board The pieces below and to the right of the chosen cell disappear after every turn Game Description Geometric 2-D Chomp: Two-player game Players take turns choosing a square box from an m x n board The pieces below and to the right of the chosen cell disappear after every turn Player 1 makes a move Game Description Geometric 2-D Chomp: Two-player game Players take turns choosing a square box from an m x n board The pieces below and to the right of the chosen cell disappear after every turn Player 2 makes a move Game Description Geometric 2-D Chomp: Two-player game Players take turns choosing a square box from an m x n board The pieces below and to the right of the chosen cell disappear after every turn Player 1 makes a move Game Description Geometric 2-D Chomp: Two-player game Players take turns choosing a square box from an m x n board The pieces below and to the right of the chosen cell disappear after every turn Player 2 makes a move Game Description Geometric 2-D Chomp: Two-player game Players take turns choosing a square box from an m x n board The pieces below and to the right of the chosen cell disappear after every turn Player 1 makes a move Game Description Geometric 2-D Chomp: Two-player game Players take turns x choosing a square box from an m x n board The pieces below and to the right of the chosen cell disappear after every turn Player 2 loses! Game Description Numeric 2-D Chomp: Players take turns choosing a divisor of a given natural number, N They are not allowed to choose a multiple of a previously chosen divisor The player to choose 1 loses 20 21 22 23 24 30 1 2 4 8 16 31 3 6 12 24 48 32 9 18 36 72 144 33 27 54 108 216 432 N = 24 * 33 = 432 Game Description Numeric 2-D Chomp: Players take turns choosing a divisor of a given natural number, N They are not allowed to choose a multiple of a previously chosen divisor The player to choose 1 loses 20 21 22 23 24 30 1 2 4 8 16 31 3 6 12 24 48 32 9 18 36 72 144 33 27 54 108 216 432 N = 24 * 33 = 432 Player 1 chooses 12 Game Description Numeric 2-D Chomp: Players take turns choosing a divisor of a given natural number, N They are not allowed to choose a multiple of a previously chosen divisor The player to choose 1 loses 20 21 22 23 24 30 1 2 4 8 16 31 3 6 12 24 48 32 9 18 36 72 144 33 27 54 108 216 432 N = 24 * 33 = 432 Player 2 chooses 9 Game Description Numeric 2-D Chomp: Players take turns choosing a divisor of a given natural number, N They are not allowed to choose a multiple of a previously chosen divisor The player to choose 1 loses 20 21 22 23 24 30 1 2 4 8 16 31 3 6 12 24 48 32 9 18 36 72 144 33 27 54 108 216 432 N = 24 * 33 = 432 Player 1 chooses 8 Game Description Numeric 2-D Chomp: Players take turns choosing a divisor of a given natural number, N They are not allowed to choose a multiple of a previously chosen divisor The player to choose 1 loses 20 21 22 23 24 30 1 2 4 8 16 31 3 6 12 24 48 32 9 18 36 72 144 33 27 54 108 216 432 N = 24 * 33 = 432 Player 2 chooses 2 Game Description Numeric 2-D Chomp: Players take turns choosing a divisor of a given natural number, N They are not allowed to choose a multiple of a previously chosen divisor The player to choose 1 loses 20 21 22 23 24 30 1 2 4 8 16 31 3 6 12 24 48 32 9 18 36 72 144 33 27 54 108 216 432 N = 24 * 33 = 432 Player 1 chooses 3 Game Description Numeric 2-D Chomp: Players take turns choosing a divisor of a given natural number, N They are not allowed to choose a multiple of a previously chosen divisor The player to choose 1 loses 20 21 22 23 24 30 1 2 4 8 16 31 3 6 12 24 48 32 9 18 36 72 144 33 27 54 108 216 432 N = 24 * 33 = 432 Player 2 loses! Fair or Unfair? Strategy-Stealing Argument Suppose player one begins by removing the bottom right-most piece Fair or Unfair? Strategy-Stealing Argument Suppose player one begins by removing the bottom right-most piece If that move is a winning move, then player one wins Fair or Unfair? Strategy-Stealing Argument If it is a losing move, player two has a good countermove and player two wins Fair or Unfair? Strategy-Stealing Argument If it is a losing move, player two has a good countermove and player two wins But player one could have gotten to that countermove from the very beginning Therefore, player one has the winning move and can always win, if he/she plays perfectly Known Special Cases m x m Chomp 1 2 Player one chomps the piece 1 located at (2,2) 2 3 4 x 3 4 Known Special Cases m x m Chomp Player one chomps the piece located at (2,2) The board is left as an L- shape, and player one copies player two’s moves symmetrically Player 1 chooses (2,2) Known Special Cases m x m Chomp Player one chomps the piece located at (2,2) The board is left as an L- shape, and player one copies player two’s moves symmetrically Player 2 moves Known Special Cases m x m Chomp Player one chomps the piece located at (2,2) The board is left as an L- shape, and player one copies player two’s moves symmetrically Player 1 moves symmetrically Known Special Cases m x m Chomp Player one chomps the piece located at (2,2) The board is left as an L- shape, and player one copies player two’s moves symmetrically Player 2 moves Known Special Cases m x m Chomp Player one chomps the piece located at (2,2) The board is left as an L- shape, and player one copies player two’s moves symmetrically Player 1 moves symmetrically Known Special Cases m x m Chomp Player one chomps the piece located at (2,2) x The board is left as an L- shape, and player one copies player two’s moves symmetrically Player 2 loses! Known Special Cases Two-Rowed Chomp Proposition 0: (a, a-1) is P-position , where a 1 (a,b) is an N-position ONLY when a b 0 and a≠ b+1 Winning Moves: (a,a-1) if a=b (b+1,b) if a b+2 Three-Rowed Chomp Zeilberger’s “Chomp3Rows” |--------a--------| Doron Zeilberger developed a program that computed P-positions for 3-rowed Chomp for c≤ 115 We will be using this notation throughout |---b---| |--------c--------| [c, b, a] Three-Rowed Chomp Proposition 1: The only P-positions [c,b,a], with c = 1, are [1,1,0] and [1,0,2] N-positions with at least 6 pieces and with c = 1: [1,1,1], [1,2,0], [1,0,3+x], and [1,1+y,x] [1,1,4] i.e., [1,0,3+x] where x = 1 Winning Moves: [1,1,1], [1,2,0],[1,0,3+x] to [1,0,2] [1,1+x,y] to [1,1,0] Proposition 2: [2,b0,a0] is a P-position iff a0 = 2 Move to: [1,0,2] Our Research Adaptive Learning Program Two computers play against each other, both eventually learn to play at their best Displays : Board 1st computer’s opening winning move P-positions and their total amount Number of games played Approximation of P-positions Amount of P-Positions as a Function of N (3 Rows) 60 y = 5.6519ln(x) - 17.226 Total Number of P-Positions 50 40 30 20 10 0 0 10000 20000 30000 40000 "N" Value 50000 60000 70000 80000 Approximation of P-positions Amount of P-Positions as a Function of N (4 Rows) 180 y = 0.906x0.4431 Total Number of P-Positions 160 140 120 100 80 60 40 20 0 0 20000 40000 60000 "N" Value 80000 100000 120000 Approximation of P-Positions Amount of P-Positions as a Function of N (5 Rows) 250 Total Number of P-Positions y = 0.3951x0.5494 200 150 100 50 0 0 10000 20000 30000 40000 50000 "N" Value 60000 70000 80000 90000 Initial attempt to Analyze P-Positions Initially we decided to look at the sum of the P- positions to note obvious patterns One obvious pattern was found (the one proposed by Zeilberger) Was not much of a success due to the various possible arrangements of pieces Analyzing Opening Winning Moves Computer’s opening winning moves for 3,4, and 5 rows were analyzed One significant pattern was observed for 3-rowed Chomp, and a possible pattern was observed as well No clear patterns were found for 4 and 5-rowed Chomp P-Positions after Computer Learned Opening Move 3-Rows Board Size: Value of N Computer 1's Opening Winning Move Resuting P-Position 3x1 9 3 [0,0,1] 3x2 18 18 [1,1,0] 3x3 36 6 [1,0,2] 3x4 72 12 [2,0,2] 3x5 144 72 [3,2,0] 3x6 288 24 [3,0,3] 3x7 576 144 [4,3,0] 3x8 1152 48 [4,0,4] 3x9 2304 576 [6,3,0] 3x10 4608 96 [5,0,5] 3x11 9216 192 [6,0,5] 3x12 18432 2304 [8,4,0] 3x13 36864 384 [7,0,6] 3x14 73728 9216 [10,4,0] 3x15 147456 768 [8,0,7] Type 1: y = [y [1]+4n,0,y [3]+3n] n 0 0 Opening Winning Move Conjecture for 3-Rowed Chomp Suppose xn is the set of board sizes: Board Size (x0): Computer 1's Opening Winning Move: Resulting P-Positions (y0): 3x1 3 [0,0,1] 3x3 6 [1,0,2] 3x4 12 [2,0,2] 3x6 24 [3,0,3] 3 x (1+7n), 3 x (3+7n), 3 x (4+7n), 3 x (6+7n), where n≥0. Then the computer’s opening winning moves for xn are to the set of P-positions yn yn has a pattern such that: yn = [y0[1]+4n,0,y0[3]+3n], where : Board Size (x1): Computer 1's Opening Winning Move: Resulting P-Positions (y1): [0,0,1] 3x8 48 [4,0,4] 3x10 96 [5,0,5] 3x11 192 [6,0,5] 3x13 384 [7,0,6] y0 = { [1,0,2] [2,0,2] [3,0,3] } Type 1: y = [y [1]+4n,0,y [3]+3n] n 0 Type 2: In Progress 0 Board Size (x0): Computer 1's Opening Winning Move: Resulting P-Positions (y0): 3x1 3 [0,0,1] 3x3 6 [1,0,2] 3x4 12 [2,0,2] 3x6 24 [3,0,3] Board Size (x1): Computer 1's Opening Winning Move: Resulting P-Positions (y1): 3x8 48 [4,0,4] 3x10 96 [5,0,5] 3x11 192 [6,0,5] 3x13 384 [7,0,6] Board Size: Computer 1's Opening Winning Resuting P-Position Move 3x2 18 [1,1,0] 3x5 72 [3,2,0] 3x7 144 [4,3,0] 3x9 576 [6,3,0] 3x12 2304 [8,4,0] 3x14 9216 [10,4,0] No Patterns Found 4-Rows 5-Rows Board Size Value of N Computer 1's Opening Winning Move Resuting PPosition 4 x1 27 3 [0,0,0,1] 4x2 54 54 [1,1,0,0] 4x3 108 18 [1,0,2,0] 4x4 216 6 [1,0,0,3] 4x5 432 36 4x6 864 4x7 Board Size Value of N Computer 1's Opening Winning Move Resuting PPosition 5x1 81 3 [0,0,0,0,1] 5x2 162 162 [1,1,0,0,0] 5x3 324 108 [2,0,1,0,0] [2,0,3,0] 5x4 648 36 [2,0,0,0,2] 12 [2,0,0,4] 5x5 1296 6 [1,0,0,0,4] 1728 72 [3,0,3,0] 5x6 2592 72 [3,0,0,3,0] 4x8 3456 24 [3,0,0,4] 5x7 5184 12 [2,0,0,0,5] 4x9 6912 3456 [7,2,0,0] 4x10 13824 288 [5,0,5,0] 5x8 10368 5184 [7,1,0,0,0] 4x11 27648 48 [4,0,0,7] 5x9 20736 288 [5,0,0,4,0] 4x12 55296 96 [5,0,0,7] 5x10 41472 24 [3,0,0,0,7] 4x13 110592 576 [6,0,7,0] 5x11 82944 41472 [9,2,0,0,0] Analyzing All P-positions by Grouping 3, 4, and 5-rowed Chomp was analyzed The P-positions within these n-rowed Chomp sets were grouped by the amount of pieces in the bottom row 4 Rows: d = 2 [2,0,0,4] [2,0,0,4] [2,0,0,4] [2,0,0,4] [2,0,0,4] [2,0,0,4] [2,0,0,4] [2,1,0,2] [2,1,1,5] [2,1,2,2] [2,1,3,2] [2,1,4,3] [2,1,5,3] [2,1,6,3] [2,2,1,3] [2,2,1,3] [2,2,1,3] [2,2,1,3] [2,2,1,3] [2,2,1,3] [2,2,1,3] [2,3,0,4] [2,3,0,4] [2,3,0,4] [2,3,0,4] [2,3,0,4] [2,3,0,4] [2,3,0,4] The P-positions for each group were then sorted into their possible permutations [2,4,0,6] [2,4,0,6] [2,4,0,6] [2,4,0,6] [2,4,0,6] [2,4,0,6] [2,4,0,6] [2,5,1,5] [2,5,1,5] [2,5,1,5] [2,5,1,5] [2,5,1,5] [2,5,1,5] [2,5,1,5] [2,6,3,2] [2,6,3,2] [2,6,3,2] [2,6,3,2] [2,6,3,2] [2,6,3,2] [2,6,3,2] [2,7,4,0] [2,7,4,0] [2,7,4,0] [2,7,4,0] [2,7,4,0] [2,7,4,0] [2,7,4,0] Pattern Found After Grouping Constant Row Value Conjecture For n-rowed Chomp, when n≥3, at least one subset of its total P-positions will have a pattern as follows: n-2 columns of the data for the subset will be fixed to distinct constant values In the following column the values will increase by a value of one The values of the remaining columns may vary or have a constant value as well 5-Rows: e = 1 3-Rows: c = 4 4-Rows: d = 4 [e,d,c,b,a] [d,c,b,a] [c,b,a] [1,0,0,0,4] [4,0,0,7] [1,0,0,1,2] [4,0,4] [1,0,0,2,3] [4,0,1,5] [1,0,0,3,3] [4,1,4] [4,0,2,7] [4,2,4] [1,0,0,4,3] [1,0,0,5,3] [4,0,3,2] [1,0,0,6,3] [4,3,0] [4,0,4,4] [1,0,0,7,3] Concluding Remarks Developed a learning program to analyze Chomp Approximated amount of P-positions per board size Initially analyzed sum of P-positions to find patterns Analyzed Computer’s opening moves and resulting P-positions Opening Winning Move Conjecture for 3-Rowed Chomp Grouped P-positions of certain board sizes with fixed boards by amount of pieces in bottom row Constant Row Value Conjecture Aknowledgements iCAMP Program Faculty Advisor: Dr. Eichhorn Robert Campbell Game Theory fellow researchers