A few years ago I worked on the AI for a WWII Strategy game called Patton Drives East.
A simpler game than Elemental, with fixed opponents and a single map (covering USA, the Atlantic, Europe and the Middle East) it utilized a personality driven AI that adopted a randomly selected strategy at game start based on random seed(s) generated for the AI player.
Example: An Allied Human player might be facing an Axis AI that adopted;
01-20: English First Strategy
21-50: Russia First Strategy
51-80: Middle East First, then Russia Strategy
---------51-60 Middle East Via Turkey
----------------------51-60 Middle East via Turkey Diplomatic Strategy
----------------------51-60 Middle East via Turkey Production Strategy
----------------------51-60 Middle East via Turkey Conquest Strategy
---------61-75 Middle East via Libya
---------76-80 Middle East via Spain (i.e. Take Spain and Gibraltar when Attack Middle East)
81-00: USA Invasion Strategy (yes, in this game the Axis could invade the USA!)
---------81-90 UK then USA
---------91-00 Spain then USA
----------------91-00 Diplomatic Strategy (diplomacy on Spain, keep Russia Neutral)
----------------91-00 Production Stategy
----------------91-95 USA Invasion Strategy 1 (conquer Spain, then invade northern USA, feint to South USA)
----------------96-00 USA Invasion Strategy 2 (conquer Spain, then invade southern USA, feint to North USA)
In each game the player faced an opponent that could be pursuing a number of different corrordinated (war, production, diplomatic) war strategies. (PS: The AI in Patton Drives East are in a text file that is easyily modifiable - and the war fighting, production and diplomatic strategies can be linked). In PDE you didn't know which front the AI would focus on.
I would like to see each Elemental AI opponent have a chance to adopt a pesonality drivien strategy (based on the strategies used by the best players), in lieu of following a general AI strategy. I don't want to see all the AI players following the same stategy to win.
You might have one AI spam cities, while another builds up a single powerful city. One AI might build an army of peasants warmed with warstaffs while the other build an army of plate-mail armored warriors. Another AI might focus on recruiting demonic armies or powerful champions. One might focus on magic. I want to face AI opponents that are smart and really different from each other in how they choose to advance their kingdom or empire. At the same time I want them to use strategy coordinated research, production, and war fighting strategies that maximize their chance for success.
