Is it a convention that one plays a game to win? More exactly, that one plays a game aiming, or feigning to aim, either to win, or, if that proves unattainable, to avoid losing? Hardly: it follows from the meanings of ‘play a game’, ‘win’ and ‘lose’. It is not a convention that there are seven days in a week, because a period of time would not be a week unless it was just seven days long. It is a convention, however, that we use the week as a salient unit of time, have names for the days of the week and use them to refer to particular days, and so forth. Well, is it a convention that in chess one aims, if possible, to checkmate one’s opponent? Certainly it is, in a sense of ‘convention’ that covers the rules of the game (and not in that which games players contrast rules and conventions). For it is as much one of the rules of chess that one wins only by checkmating one’s opponent as it is that the Rook cannot jump over an intervening piece. In Chinese chess, for example, one can win by putting one’s opponent into stalemate: the rule is different. Now if it follows from the meaning of the word ‘win’ that one plays a game with the real or apparent aim of winning if one can, it must be a rule of chess — one of the conventions constitutive of the game — that one play it with the real or apparent aim of checkmating one’s opponent. Of course, the rules, in their standard formulation, do not state this in that form; rather, they use the word ‘win’ and rely on our understanding of it, which involves our understanding of the general practice of playing games.