minimax - How this evaluation function work in a Connect 4 game? (Java) -

i exploring how minimax algorithm can used in connect 4 game alpha-beta pruning.

so looking through source code connect4 player strategy , found evaluation function:

/** * score of board */ public int score(){     int score = 0;     (int r= 0; r < rows; r++) {         if (r <= rows-4) {             (int c = 0; c < cols; c++) {                 score += score(r, c);             }         } else {             (int c = 0; c <= cols-4; c++) {                 score += score(r, c);             }         }     }     return score; }   /** * helper method score of board */ public int score(int row, int col){     int score = 0;     boolean unblocked = true;     int tally = 0;     //int r, c;     if (row < rows-3) {         //check         unblocked = true;         tally = 0;          (int r=row; r<row+4; r++) {             if (board[r][col] == checkers[1-playertomovenum]) {                 unblocked = false;             }             if (board[r][col] == checkers[playertomovenum]) {                 tally ++;             }         }         if (unblocked == true) {             score = score + (tally*tally*tally*tally);         }         if (col < cols-3) {             //check , right             unblocked = true;             tally = 0;             (int r=row, c=col; r<row+4; r++, c++) {                 if (board[r][c] == checkers[1-playertomovenum]) {                     unblocked = false;                 }                 if (board[r][c] == checkers[playertomovenum]) {                     tally ++;                 }             }             if (unblocked == true) {                 score = score + (tally*tally*tally*tally);             }         }     }     if (col < cols-3) {         //check right         unblocked = true;         tally = 0;         (int c=col; c<col+4; c++) {             if (board[row][c] == checkers[1-playertomovenum]) {                 unblocked = false;             }             if (board[row][c] == checkers[playertomovenum]) {                 tally ++;             }         }         if (unblocked == true) {             score = score + (tally*tally*tally*tally);         }         if (row > 2) {             //check down , right             unblocked = true;             tally = 0;             (int r=row, c=col; c<col+4; r--, c++) {                 if (board[r][c] == checkers[1-playertomovenum]) {                     unblocked = false;                 }                 if (board[r][c] == checkers[playertomovenum]) {                     tally ++;                 }             }             if (unblocked == true) {                 score = score + (tally*tally*tally*tally);             }         }     }     return score; } 

i found code in pdf: http://ryanmaguiremusic.com/media_files/pdf/connectfoursource.pdf

i want understand how evaluation function works , decides best move made... give me help? appreciated.

here general answer:

the evaluation should give better values better positions. in games position evaluated calculating score following way: increase score desirable configurations/events , decrease undesirable ones. deciding how evaluated feature should change value (= balancing weights) can difficult.

if apply connect four 1 feature number of threats alive. algorithm (which solves 7x6) have if winning move on odd or line. , there rules "if 2nd player has 2 threats has won game" (it comes down when filling board , moves forced, rule given little simplified: 2nd player have kill threat if can't fill other columns if 1st player has odd threat there).

simple example rule given (o = first player, x = 2nd player), x wins:

  * x x x * o   o x o   o 1 2 3 4 5 6 7 

there once detailed scientific paper it. closest find @ moment: http://web.mit.edu/sp.268/www/2010/connectfourslides.pdf should give ideas.

btw. opening books (generally predefined wisdom in form e.g. joseki, fuseki) , special end game evaluators can improve performance of minimax.