The AI build priorities don't always match the human build priorities. This is expected, as it is often necessary to compensate for the AI's poor military tactics. However, even taking into account the differences between AI and human optimal strategies, there are some areas where the AI build priorities could be improved:
With the above in mind, it would appear that the 'build-often' items can be changed for each civilization, resulting in better AI build priorities.
Another way of influencing build priorities appeared to be through the use of flavors. Unfortunately, flavors don't seem to influence the AI that way at all. Until we gain a better understanding of flavors, the AU mod will not use them to affect build priorities.
So what do you think? Does the above proposal for build-often items seem reasonable? Do you have any other ideas of how to help the AI in its build decisions?
- The AI uses its workers inefficiently and takes a long time to improve its land. The AI would clearly benefit from more workers.
- The AI does not use the luxury slider, so it runs into significant happiness problems, especially in large cities building Wonders. The AI would benefit from building more happiness improvements and Workers.
- The AI often misses trade opportunities because it delays building harbors, and even when it trades luxuries, it does not benefit from them as much as it could because it doesn't have enough marketplaces. Given the AI's happiness problems, it would clearly benefit from building more trade improvements.
- The AI often delays building factories, and as a result its Industrial Age production suffers. The AI would benefit from building production improvements.
- The AI often fails to take advantage of its traits. Scientific civilizations would benefit by building scientific buildings at a higher priority, Religious civilizations by building happiness improvements, Militaristic by building militaristic improvements and units, et cetera.
- The AI cities often grow too fast for their own good. Since the AI doesn't know how to set up Settler and Worker factories, and since it has happiness problems, the AI would benefit from building fewer growth buildings.
- The AI does not build enough air units, now that airpower has been strengthened in C3C. See the Balancing Airpower thread.
With the above in mind, it would appear that the 'build-often' items can be changed for each civilization, resulting in better AI build priorities.
- All civilizations should have Workers, Trade improvements, and Air units as build-often. This will encourage the AI to build Workers, Marketplaces, Harbors, and planes.
- All except Scientific civilizations should have happiness improvements as a build-often item. This will help the AI build Temples, Cathedrals, and Marketplaces.
- Scientific civilizations should have Cultural improvements as build-often, instead of Happiness improvements. This will help keep the number of priorities down. Scientific improvements have more culture than happiness improvements of the same level.
- Militaristic civilizations should have offensive units as a build-often item. Defensive units are already built often enough by the AI.
- Seafaring civilizations should have naval units as a build-often item, in order to take full advantage of the trait's benefits.
- Civilizations with 'builder traits' should have production as a build-often item. It's not that civilizations that need to build units do not need the extra production from factories, but keeping the number of build-often items low should make each build-oftem item more effective. Aggressive militaristic civilizations should usually not be building a Courthouse instead of another Swordsman.
- Growth improvements should be removed from all civilizations' build-often lists.
Another way of influencing build priorities appeared to be through the use of flavors. Unfortunately, flavors don't seem to influence the AI that way at all. Until we gain a better understanding of flavors, the AU mod will not use them to affect build priorities.
So what do you think? Does the above proposal for build-often items seem reasonable? Do you have any other ideas of how to help the AI in its build decisions?
Comment