Overspecialization :  Experiment 2                        (click on picture below to see simulation)

Overspecialization:  The selection pressure to outcompete one another can help ensure that any weaknesses in a player are discovered and the individual destroyed. However, the ability to focus on the opponent's weakness can provide an easy way to win. This may produce degenerate players that over-specialize on opponents weaknesses, and fail to learn a task in a general way. For example, I develop my pawn structure skills to beat you; you develop knight/bishop skills to beat me; I develop rook skills to beat you; etc.  My learning pawn structure might lead to being better than you, but if I've been neglecting all my other skills they may be forgotten and in the long term I have not necessarily become an objectively better player in general.

In this experiment, for each game between 'a' and an opponent, determine D, the dimension or skill in which they are most different. The winner is the player who is greater in dimension D. i.e. If a[D] > Opponent(i)[D] then 'a' gets one fitness point, otherwise fitness remains unchanged. So in this game only one skill is relevant when playing against a particular player - but which skill matters depends on who you play. However, in this game, a change that increases your score against one player never decreases your score against another player - this makes the game transitive. Look for the fact that fitnesses do not climb as high as they do in experiment 1a where players only have one skill. This is caused by the 'atrophication' of skills that are not currently being used.

Relativism or Intransitivity occurs when there is no clear direction of improvement for a skill, and the only point of reference is the opponent. 


Home  |  Paper  |   Simulation   |  Experiments:   1a   1b    3