I am beginning to think I should set modding the AI as my new challenge. Since C3 has become unplayable as the client if you like big unit games and live in the Antipodes with high internet latency, then I might as well try to mod the AI so that it gives me competitive games. All other C3 hopes seem... well hopeless, given the current state of affairs. Which are the best current AI mods around and what do they do? What can they do? AI modders in particular please comment. I believe a non-cheating AI could be made very, very good. All one has to do is harness the AI's micro-ing ability intelligently. In theory, this should not be too hard. In practice it would be a lot of work. Module 1. - Resourcing and Economic Growth System The first step would be to make a mod designed to only run the economy. The idea would be to optimize all the decisions for economic growth. There would be no military component. One would play against this AI as an economic competition only. The winning goal would be the greater economic score after a set peace time ends. The goal would be most peasants, most economic buildings and most resources. Once the AI can grow faster economically than any human can do then this module is finished. The key to getting fastest growth would be solid heuristics guiding some basic look-ahead tree search algorithms which checked theoretical growth for the next x seconds from certain possibly branching decisions. C3's inflation model of costs actually self-prunes this tree search to really quite small dimensions. Module 2. - Military Growth System - Add the stage 2 system which would convert a portion of economic growth and upgrades to military growth and upgrades. The AI would need heuristics and algorithms to calculate by how much and when to start converting economic growth to military growth and upgrades. It would use the same routines to estimate what the opponent is doing on the assumption that the opponent is as agressive and micro-ing as the AI (is going to be). Again, the C3 design would aid this goal. A solid heuristic for military growth is to get each military building as soon as possible in the possibility tree search and then set that building to infinite production. Military upgrades are more complex. The heuristic here is to avoid them as long as possible and then rush them as fast as possible if facing an attack or launching an attack. Module 3 - Military Tactics System. This would be complex and would involve tactics for engagements including combined arms action tactics (infantry, cavalry, artillery). Best to write little here or this segment will get too long. Module 4 - Strategy System This module would link and command the other three modules, passing commands and requirements to and between the other modules. Also, it would generate a full military strategy based on these principles; (1) Internal lines defense - Gaining a wide field of view by micro-scouting (the AI will not cheat) and then concentrating forces to repel attacks at good defensive points with superior numbers. (2) Excellent micro-scouting for all weaknesses and to detect incoming attacks. (See 1.) (3) Concentration of force at the decisive points (decisive engagements). (4) Economy of force to raid and to guard its own economy. (Just enough forces for these tasks and no more.) (5) Counting attrition rates - The AI will calculate accumulated attrition rates on both sides, estimate most likely enemy production levels from all intel, and then assess when it has the enemy attrited enough to launch a major damaging or decisive attack. Summing Up In theory, the above done properly will defeat almost any human opponent except the most strategically adept and patient persons and it will do so without cheating. Iteratively weaker levels would have to be made to give most players a chance.
