What are some decent 4X / Strategy / Grand-Strat games that have actually decent, competent AIs to play against? It seems as if though every single one I try out has potato-tier AIs that can just be easily exploited through loopholes / doing cheeky things the devs never thought of.
Besides games like Chess and recently Go as well, no, not really. There are some games where the AI is decent (it's a tactics game but Ultimate General: Gettysburg actually has a kinda competitive AI for example: the difficulty system is also very neat since it doesn't have artificial bonii but instead difficulty comes from personality traits gone too far, like easy aggressive AI being reckless while hard aggressive AI is decisive and determined but not suicidally so) but even they are quite few and far between from what I've seen. In so many games the AI plays effectively by random and its shittiness is only constrained by game mechanics not giving it enough rope to hang itself with.
>>343872757
I have this sinking feeling that AIs are inherently handicapped by how complex a game's rule-set is, and that the only reason games like Chess and Go have decent AIs is because their rule-sets are extremely simple and understandable from a computer-AI point of view.
Am I right?
shogun 2 total war has pretty decent ai
>>343871720
Galactic Civilizations 2 has one of the best AIs in any game ever. It's not unusual to see it play better than humans.
>>343873674
I made this thread after being disappointed by Gal Civ 2's AI, actually. One thing that stood out to me was how the AI would always suicidally rush space-stations you build, even ones that have absolutely no modules installed. Its very easy to bait the entire AI's fleet to go where you want them to by just building clusters of cheap space stations because of how overwhelmingly the AI prioritizes destroying them above all other things, regardless of if they even have modules installed or not.
>>343873863
hmm, which version of the game do you have?
>>343874016
Steam version.
>>343872895
Depends on how you look at it.
Computers are good at Chess because Chess is a game of complete information and it's easy to make a tree search of all possible moves. However, even in Deep Blue had some darn complicated evaluation systems programmed in (for example, it's generally understood that a bishop has value of 3 pawns and so does knight but bishop pair is more valuable than a knight pair, and Deep Blue didn't just calculate every possible move but took hundreds of such rules into account and modern Chess engines are even more sophisticated) so it's not just brute force. A lot of research and hundreds of thousands of man-hours has been spent on getting Chess AI to the point where it's today (untouchable by humans).
Go is a different beast because, while it's also a complete information game and it wouldn't be difficult to write an algorithm that goes through all possible moves, that gets you nowhere because of astronomically high number of moves (game tree complexity is 10^360) and greater reliance on long-term goals rather than gaining material in small-scale skirmishes. About ten years ago computers were total garbage at Go but then they started seeing some level of success (beating good amateur players) thanks to Monte Carlo tree search. However, even Monte Carlo -based programs lacked "true" understanding of the game and had weaknesses that a human player could easily exploit (like being incapable of correctly handling Ko fights). It wasn't until last year when Monte Carlo method was combined with deep neural networks (something akin to how humans intuitively evaluate board positions) and AlphaGo beat top-level professional Lee Sedol earlier this year.
So both those games have characteristics that make them easy for computers to evaluate (no fog of war, very clear and precise rules and such like) but ultimately the big difference-maker is that they have a lot of effort put into them over the decades while most developers don't even try.
>>343874040
The ultimate edition, with the expansions, that's what I mean?
I don't remember that happening, but yeah, it sounds pretty shitty
>>343874434
Yeah, its ultimate edition. Its actually kind of amazing - because not only will the AI overwhelmingly prioritize attacking space-stations above even your own fleets / constructors / asteroid mines, but at the same time they'll just go full potato and spam-attack very heavily defended space stations with trickles of single-ship fleets rather than consolidating them all into a single, massive fleet that might actually have a fighting chance at taking down the station.
Besides that, though, playing against Bright level AI can be interesting - they seem to have perfectly optimized build orders (very early industry focus / colony ship focus) and so on. Its such a shame that the AI falls apart in the field of actual combat and strategic planning.
