Now, that''s a better question. Too bad there is no easy answer.

To summarize, you should :
- model the concept of tactics/strategy for your game (I assume it is a strategy game)
- adapt the weights of sub-strategies to the player

In the (LGPL) AI Toolkit I''m working on, typically, this means :
- reduce the game to a set of elementary rules
- let the toolkit combine them into more complex rules/strategies
- apply Evolution Algorithms on the set of rules/strategies to have several competitor models of the player
- select them and make them evolve into an agent hopefully able to mimick the player

Of course, there might be better ideas. I would be willing to read them.

Be reading you,
David