The preceding code satisfies the tests, but is not necessarily the final version. It served its purpose of getting code coverage as quickly as possible. Now, since we have tests that guarantee the integrity of the expected behavior, we can refactor the code:

private static final int SIZE = 3;

public String play(int x, int y) {
  checkAxis(x);
  checkAxis(y);
  lastPlayer = nextPlayer();
  setBox(x, y, lastPlayer);
  if (isWin()) {
    return lastPlayer + " is the winner";
  }
  return "No winner";
}

private boolean isWin() {
  for (int i = 0; i < SIZE; i++) {
    if (board[0][i] + board[1][i] + board[2][i] == (lastPlayer * SIZE)) {
      return true;
    }
  }
  return false;
}

This refactored solution looks better. The play method keeps being short and easy to understand. Winning logic is moved to a separate method. Not only have we kept the play method's purpose clear, but this separation also allows us to grow the winning condition's code in separation from the rest.