In a real-time strategy game on sufficiently large scale that the AI's tactics matter it's almost always possible to find a consistent weakness in the computer's game-play. The fact is computers just don't have the power to ante up to a brain in this area. This deficit can be partially covered with the following approach:

Keep a record of battles won vs lost against the player and the odds at the start. Keep a record of various tactics used by the computer against that player. Randomly modify those tactics keeping the most successful over time.

When a series of tactics is successful against many players start with that series.
-- Voice, Jan 21 2011

Remember this step http://xkcd.com/534/
Just saying. [MechE, Jan 21 2011]

Only if they also allow me to teach my units proper tactics at the same time. The inability to get some units to fall back under appropriate conditions is a major weakness of this type of game.
-- MechE, Jan 21 2011


Other approaches to help out loser AIs which are possible in virtual world where anything goes:

1. Incite rebellion. Player's forces start fighting each other. Player can be notified that AI has incited rebellion.

2. Large holes open up below players forces and they fall in. This worked well in the movie "Return of the King" and so should be doable in other completely fake scenarios.

3. AI forces can fail to die when hit. Player can be notified (or not) that opposing forces have "manned up" with suggestion that player stops complaining and does the same.

4. Fires. AI causes players' forces to catch on fire and pop. This is made possible because the AI is friendly with the simulation program and they have started seeing each other after work.

5. Augment AI's resources by several orders of magnitude. If still inadequate to win, augment by several more orders of magnitude.
-- bungston, Jan 21 2011


Having played computer games for over a quarter of a century (no exaggeration), I have to say that the most disappointing aspect of them is the lack of advancement in the AI. As bungston so eloquently illustrates, many approaches have been adopted (in more or less obvious forms) in order to cover up for this lack of progress. The one he missed out though is what has become the most common approach, namely "Ooh, look at the shiny graphics!". I therefore vote + for this idea.
-- DrBob, Jan 21 2011



random, halfbakery