computer-go: Meta-strategy (was: A problem with understanding lookahead)

["Vlad Dumitrescu" <vladdu@xxxxxxxxxxx>]

Hi

There is an aspect that bothers me... it applies to all computer games. 

One of the big differences between computer players and human players is that computers are much, much less flexible and even if they learn, it's not as fast as a human. If the opponent's identity would be unknown and/or only one game is played, then I think "stronger" computer players could be written. By stronger I mean with a larger percentage of wins. 

That is because the human player could not choose a strategy known to work well against this or that opponent; and he/she could not devise a strategy from the first 2-3 games that will allow winning the next 7. Conversely, the computer program cannot easily shift playing strategy in order to avoid being "read".

It is a very important aspect, I think. And very difficult to build into a program. Everyone has strengths and weaknesses, and humans do use that kind of knowledge about the opponent while at the same time trying to make it difficult for the opponent to apply the same to them. 

Any thoughts about that?
/Vlad