Just to refresh your memory, here’s what a certain person had said about Empire Total War:
“Commonly, there’s this curve where I enjoy a game, and then I master the system, and then - unless it’s got a good AI - I lose all interest because I realize that mastering the system is where the challenge ends. Once I reach that point, the game is dead for me, and I hate that! That’s when the game should really start to take off.”
I’m a little tired of strategy games being paper tigers. Games like Endless Space 2 amaze me with great interesting mechanics, but then it turns out that maximum difficulty AI is unable to do anything against me even before I learn all of the game’s rules. All those intricacies, unique factions, and intrigue are lost, they’re not needed - even though learning the game was great fun. Same for Civilization 6, Hearts of Iron 4, Stellaris - and in those games this downward ride comes even sooner.
The common solution I see for strategy games is to actually end the game as soon as “tutorial” ends. XCOM or Slay the Spire or even Into the Breach unlock new things for you to see up until the final confrontation. There are achievements for those who want to walk an extra mile but those often seem like a vanity award, normal players aren’t supposed to play it.
So what modern single player strategy games are playable after you learn the mechanics?
For theme, Alpha Centauri. But much as I love it, it is dated and it does have weird bugs and rules you can exploit in ways that probably weren’t intended, and the AI isn’t great.
Offworld Trading Company has a pretty decent AI and a single-player campaign I enjoy. The campaign injects a bit of asymmetry that is much needed: it’s okay the AI players can’t compete on a fair footing, because the footing is anything but fair.
Both Infested Planet and Invisible Inc takes the idea of asymmetry to the extreme. These are great, interesting strategy games that are totally different, and have a board-game-like approach to the the clear rules. You are designed explicitly as the only ‘player’.
Stars in Shadow has very deliberately taken a single-player approach to design. It is not a perfect game by any means, and in some ways it’s too simple. The AI is okay but lacking here and there (still a million miles ahead of Endless Space 2 or most other TBS games, though). Development is continuing, but it’s a two-man-band and so taking time. I’m really hopeful this will continue to grow, though.
But when all is said and done, I cannot think of a single game that is complicated enough to feel ‘grand’, full of interesting mechanics and cool design, yet also with an AI (and supporting design for the AI) that provides interest and challenge. I am moving more and more toward expressly asymmetric, single-player design games for this reason. I am just hoping one with a bit more complexity shows up some time. (Slay the Spire is actually pretty close.)
Unless I’m misunderstanding your point somehow, I disagree with that characterization for both Into the Breach and Slay the Spire. Both of those games have clearly been built for the normal player to play dozens of games to a conclusion, unlocking new content between the runs. These unlocks aren’t there for the vanity of hardcore players, they’re for allowing the developers to ramp up the complexity of the content in a controlled manner. Start the player off with a fairly plain character/deck/squad, and then make it more complex as the player gets more experience.
The point with Into the Breach and Slay the Spire is that they’re all about extending the left part of Chick Parabola. They’re all about learning. You unlock stuff for most of the game. There are dozens of hours between starting the game and unlocking everything and then once you unlock the last thing that is there to unlock you’re supposed to be mostly finished with the game. Slay the Spire expands this time with Daily runs, but still there’s a clear roadmap: you learn basic stuff and unlock cards; you learn what is there in the game to complete act 3; you do the same with 2 more character; you unlock new short act and some additional mechanics on your way there; you know how this new small act works - and bam, the game is completed shortly after you see the last unlocked boss for the first time.
It’s all about learning new stuff. Compare it to, say, older game like Heroes 3 where people use the same units for decades and are still having fun.
I mean, a competent and sufficiently variable AI that remains interesting to play against in a complex symmetrical system like a 4X is currently a multi million dollar research project with crazy compute requirements, at best. It just doesn’t exist, and won’t, until a few more magnitudes of desktop compute power (or much closer to full interconnection to shared supermassive compute resources).
I would say that some games do some things pretty well. Pandora does economic growth well. Imperialism 2 does expansion and econ well. Age of Wonders 3 does tactical combat well. All three are games I have thousands of hours with, and even in those exceptional areas the AI is quite predictable.
I think the answer you’re looking for is Diplomacy against humans ;)
More importantly, crazy R&D time and money. You’d have to hire a few specialists in the subject and give them free reign at some mid-point in development. And even then, you may not get something workable out. So you’d need a backup AI strategy.
It’s AI in much simpler, boardgame-type games - like Space Hulk, or Blood Bowl, or Armello - being so poor that I find much more unforgivable.
Yeah. But even those are pretty complex systems when you start breaking down game states and goal states and mapping actions to the deltas found therein (or whatever approach actual AI developers use; I’m an acolyte of the Omnissiah at best).
It’s interesting how people don’t complain about the XCOM AI so much. It’s obviously a quick weighted random roll against a bare few actions, with basically no overarching strategy. But who knows what the aliens are thinking anyway? And it’s so massively asymmetrical that they don’t actually need to function beyond being sufficiently in the way of the player and the next objective.
XCOM is helped more by the fact that you’re not really supposed to replay it. You discover new stuff constantly.
Answers like Invisible Inc and Infested Planet work better. Those game are asynchronous, AI is not supposed to compete with you, it’s supposed to behave dumb. And it works.
They are complex systems. But I think there is a lot of room to be gained from even simple heuristics. I think it’s mostly a function of those sorts of games having smaller dev teams.
But it just shows how hard the problem is: if you can’t even get the small board-game style games right, what hope do the grand sweeping 4X strategy game AIs have?
I kind of look at Butterfield’s Enemy Action Ardennes and its solo gameplay and i understand just how challenging it would be to create an effective AI opponent in a fixed map with fixed units and a lot of game mechanic help to make it work.
Trying to imagine it in random maps or just differrent maps with weird states, differing objectives… it’s pretty nuts that someone even gets anything done with things like 4x games.
They can hope that devs actually think about how their games would be played by AI.
It’s clear that many mechanics of games like Civilization were developed without any care for the AI - starting with 1 unit per tile idea. Meanwhile, in a game like Heroes 3 the design makes it so that even dumb AI is an interesting opponent with a little boost to his armies, and it doesn’t feel cheap.
The Total War games (including Warhammer) are very good examples of the Chick Parabola. Master a few techniques/tactics and then game the AI until you get bored with the mid-late game slog of painting a map.
That’s definitely not my experience with the Mortal Empires campaign. I have 200+ hours in the game and still have yet to really get good at more than maybe two factions, and those factions I still lose with from time to time.
I remember the AI in HOMM II being sensible in individual fights, but fairly deterministic, while the overall scenarios leaned more towards puzzle design.
I’m going to lean towards the notion that you have to have an asymmetric design like Invisible Inc or AI War for the AI to really hold up to scrutiny.
Or you have to go more chess-like, with fairly simple rules. I think Armageddon Empires held up pretty well, but that might be rose-colored glasses.
HOMM 2 and 3 had AI good enough to be speedbumps along the player’s stomping through the largely PvE maps. They’re not “good” by modern standards, no. And I say that as someone who would probably literally marry those games if it were legal and I weren’t already spoken for.
Very much so.
I did play a ton of skirmish in those games, though, and they are excellent skirmish games. The AI in my opinion is very much “good enough” since so much of the game is PvE even in skirmishes.
Total Warhammer is on legendary a bitch to play with certain races, certainly a challenge, try Orc.
Hoi 4 with expert AI mod, can be a real challenge.
Civ 6 on immortal for example gets spammy and a tad sad panda.
Best AI that will give you a licking is Sins of a Solar Empire, it is innovative and pretty solid, same goes for Ashes of the Singularity, but Sins is just better imho.
Sadly the more complex the game is, the worse it gets, human players needed in general for now.
I remember one game, Conquest of the New World, the AI was absolutely shit at it, but it had a combat minigame…damn it rocked there, it knew its tic-tac-toe.
