sparse<double> sprand(size_t nrow, size_t ncol, double density)
{
assert(density <= 1 && density > 0);
size_t n = (size_t)(density * nrow * ncol);
dense_vector<size_t> rows(n);
dense_vector<size_t> columns(n);
dense_vector<double> values(n);
for (size_t i = 0; i < n; ++i)
{
rows[i] = (size_t)RANDI(0, nrow - 1);
columns[i] = (size_t)RANDI(0, ncol - 1);
values[i] = RAND();
}
return triplet<double>(rows, columns, values, nrow, ncol, n).to_sparse();
}
move_result karl_play(player* self, state* st, move* mv) {
int N = ((karl_params*) self->params)->N;
state test_st;
move reasonable_moves[NMOVES];
int num_moves = go_get_reasonable_moves(st, reasonable_moves);
if (num_moves == 1) {
*mv = reasonable_moves[0];
return go_play_move(st, mv);
}
color me = st->nextPlayer;
color notme = (me == BLACK) ? WHITE : BLACK;
int win[NMOVES];
int lose[NMOVES];
double pwin[NMOVES];
int i;
for (i = 0; i < num_moves; ++i) {
win[i] = lose[i] = 0;
pwin[i] = 0.0;
}
// Do playouts
for (int i = 0; i < N; ++i) {
state_copy(st, &test_st);
int starting_move_idx = RANDI(0, num_moves);
move starting_move = reasonable_moves[starting_move_idx];
go_play_move(&test_st, &starting_move);
playout_result result;
go_play_out(&test_st, &result);
if (result.winner == me) {
++win[starting_move_idx];
} else if (result.winner == notme) {
++lose[starting_move_idx];
}
}
// Calculate chances
double best_pwin = 0;
move best_pwin_move;
for (i = 0; i < num_moves; ++i) {
if (win[i] + lose[i] == 0) {
pwin[i] = 0;
} else {
pwin[i] = (double) win[i] / (win[i] + lose[i]);
}
if (pwin[i] > best_pwin) {
best_pwin = pwin[i];
best_pwin_move = reasonable_moves[i];
};
}
wprintf(L"\nDecision heatmap\n");
go_print_heatmap(st, reasonable_moves, pwin, num_moves);
wprintf(L"Going with ");
move_print(&best_pwin_move);
wprintf(L" at %.1f%%\n", best_pwin*100);
*mv = best_pwin_move;
return go_play_move(st, mv);
}