Broken AI on Marathon Speed.

[Heisenberg Particulate]

I have seen, more in this game than previous versions, that the AI, even on Immortal and Deity, just don't keep up.

Has anyone else noticed this?

I prefer longer games, Standard Speed means eventually most things take 2-5 turns to complete; units, techs, civics, etc... apart from Wonders. Marathon means you can spend time traversing the map, ages last longer for better wars, etc...

But the AI eventually stalls, I've tried turning off Barbarians, as my theory is that the AI is wanting to pump huge amounts of units first and with them getting depleted, they don't focus on science/culture. I've ended up getting 350+ Science and the highest AI is on 70. That's on Immortal, Marathon, and I was Pericles. My Culture was 860+.

My last Deity game, when i hit 400+ Culture/Science, the AI was only reaching 120.

Does anyone else think the AI is crap for Marathon? Spending nearly £100+ on a game for its expansions, I don't think it's unreasonable to expect that the AI and game modes they include in the game to provide consistent challenge. I remember when this was first released when they recommended ''using standard speed as intended as that's what the AI can play for the moment, wait for updates.''

[plaguepenguin]

The AI has never been good at this game. Whether that's because Firaxis is too cheap to spring for smarter AI, or because this game is fundamentally different from chess and therefore presents greater difficulties producing a smarter AI, I don't have the expertise to answer. I can say from my non-AI expert observations though, that the AI fails most notably at the system of systems aspect of this game that makes it enjoyable for humans. The AI can be reasonably competent within a particular system, but set it to a task that requires balancing effort towards different systems, or system-to-system interactions, and it just fails completely. Even just within the combat system, it fails most notably at acting towards a sound and coherent war objective, or reaching a sound division of forces between theaters of war, even while it isn't at all bad at counter-punching. I'm not sure that a Deep Blue for Civ would be good at that system or systems problem, because chess is just one system, though Lord knows its one system is way deeper than any single system of Civ.

To compensate for the inability or unwillingness to come out with an AI that can compete in actual intelligence, the developers try to give the human player some competition from the AI players by giving the AI heavy-handed advantages. In Civ6 they seem to have gone in for giving the AI three starting settlers and a starting army as a greater component of that advantage, compared to the permanent yield advantages they also give in 6, but relied on more in earlier versions. This makes the early game on Deity challenging, but once you survive the early game, your actual intelligence as a human allows you to employ actual strategy in increasing yields, and you can soon more than outstrip the yield advantages the AIs get on Deity. One fix would be to go back to relying more on better yields, and less on starting advantages. A further refinement would see the AIs' yield advantage go up in each era, maybe starting where it is now, but then marching upward to compensate for your human intelligence advantage. That's something that should be moddable, if that's what you want. Should Firaxis do this work for you and have advanced settings for difficulty level to let you choose exactly how the AI is compensated for being not as intelligent as you are? Perhaps.

I have noticed that AI civs in 6 don't, nearly as often, get one competitor for the human civ in the late game. In 5, I noticed a marked increase in how often this pattern -- the other major land mass would have a superpower mostly take it over completely by the late game to form a worthy opponent to your civ for a late game world war -- from Prince to King. I assumed that was because the higher yield advantage to all AI civs on King was like steroids, and the civs programmed to be expansionist would succeed better at their career of conquest because they were on steroids at the higher difficulty level, and so conquered their AI neighbors who were using the yield advantage to pursue more pacific programs. Maybe this difference in Civ 6 not having a single AI civ conquer many of the others so it can give you a worthy late-game opponent arises from the developers having made AI civs more flexible in deciding on goals, in a premature and unsupportable attempt to make them more human-like. This decision on ultimate goals is the highest level system of systems decision, so of course the AIs are no good at it, and the effect is to just make them indecisive, so you less often see that some single-minded conqueror AI civ has conquered the whole other land mass in a continents game, because instead of being single-minded in a course of conquest, it foolishly pursued some other victory type in its greater flexibility.