AI Players: Based on tread: Do AI players play by the same rules as human players

To sum up the original thread in 2016 it was pointed out a gamer played against AI that had elite units and wondered how. It was answered, correct or not I know not, but assuming it is, that the AI player had accumulated persistent crate rights such as blueprints which you carry forward to future games. My response below is to that information.

Crates are intriguing. They appear to be designed to temp you into making an attack or going to war with a new nation whether it is wise or foolish. This is a pretty good simulation of real world complications in that respect, so I am for it.

AI (called Ministers) also plays for inactive players. I don't think it would be fair for this type of AI player to care over previously accumulated crate right or to inherit them from the inactive player. This is particularly dangerous when the AI controls a playable country which a previous player may have improved a lot before losing interest.

Conceivably the AI players can play many more game hours than a real player can. If allowed to accumulate crate advantages such as blueprints persisting and growing larger over a large number of games you would expect these player to eventually get very strong. Some limit should be enforced on how much they can keep. Perhaps each round some % chance of losing something should exist so that as it becomes stronger probabilities become stronger that they lose something. Determining stability to keep this within bounds is a mathematical problem.

AI algorithms actively encouraging an AI player to seek out crates can be beneficial from the standpoint of creating random AI attacks and new Wars. It just needs to lead to a stable system where AI is not dominant. AI players can be made stronger in other ways such as some finished research at the start and extra troops or buildings or even money to buy resources it does not produce. These additional benefits can be randomized to make it impossible to know how dangerous a particular AI country is. Humans will adapt to additional danger from AI by taking the risk of a nearby AI country more seriously.

My expectation is that not all versions of the AI program are identical. The adventure of experimenting with AI rules to discover what works best would be hard for me to resist if I was the programmer.