But when you break that down into just numbers, it shouldn’t be that hard. Perhaps the issue was more the human AI designers not being that good at math.
Can we play custom species?
Tee hee… ahem sorry.
The thing is that Brad has shown himself to be at least good enough at math to come up with quite good AI in the past (e.g. GCII). Not sure if he did the AI on GCIII, but I can’t imagine it’s as simple as “not good at math”.
Will we be able to build planet killers? Will there be supply routes?
Well its a trivial task for players to optimize planets, at least high tier ones. I can’t think of any issues where the algorithms for an AI to figure this out would be too hard. There’s nothing like language or temporal processing or anything social. Its essentially just adding and multiplying. Even the spatial component should be relatively simple, its a grid iirc and no z-levels or anything.
Here’s what I think would be a good question:
What is Stellaris bringing new to the table? What’s going to make it different from other galactic 4X games?
Didn’t they already answer that though? Its a galactic grand strategy game for one, and on top of that it has endgame events, a totally new tech system compared to normal tech trees, and so forth. Its all in this thread and the RPS preview.
Pretty sure the Q&A was intended to be about more specific stuff than could be put in a standard publicity packet.
Question.
I love the idea of really wild events that can change the game in more ‘aggressive’ ways than in your usual 4X game. In fact, it sounds too good to be true. What can you say to make me a believer? What concrete data? Because the cynical me thinks it’s all an exaggerated by the press and PR right now.
My guess (as programmer who hasn’t played GC3) is that the problem isn’t optimization of buildings in the “now” for the AI, that’s quite easily achived. The bigger problem is certainly planing in advance. A player will leave tiles unused for future use because he knows he will research tech X so he can build Y and put it on the tile Z. That’s where the real difficulty is for the AI. There are plenty of algorithms which can handle the turn to turn stuff (and better than the player) but it all starts to crumble once long term goals are involved because it increases complexity exponentially.
There are ways around it but yeah it’s hard and sadly AI programming is most of the time just an afterthought in many games.
e: Question for the Q/A:
How exactly will battles work? Is it a nice visualization of an abstracted system like for example the sea battles in HoI or are they regular RTS battles only without player input?
What options does the player have to influence battles?
Well the original post was about how the AI just randomly vomited down buildings. If it got initial building placement right, that seems unlikely. Plus the cheap shortcut is just to have the AI redo buildings whenever a major tech comes out.
I mean, we can have a useful debate about it over in the GC3 thread if people would like!
It was really just meant to be illustrative of AIs not knowing how to (or not needing to) play the “same” game as the player. Unlike an asymmetrical game like Sorcerer King, I suspect the goal in Stellaris is to have multiple randomized civilizations starting out on a relatively level playing field, dealing with randomized events, and eventually meeting and interacting. I’d be disappointed if the foes were, for instance, just a bunch of timer-checks. “Oh, it’s Turn 100, time to give all the AIs Destroyer-class ships every 5 turns now.” Knowing the AI is out there, dealing with and doing the same sort of stuff as me is mildly rewarding, but there’s also the fact that I want to be able to systematically take them apart/disable/cripple them in ways that are obvious to me. Knocking out all their econ starbases or infinium mining rigs or depleting their Diplomancy Point pools or whatever it might be that would hurt ME grievously should do the same to them, ideally (at least at the baseline difficulty–I understand the need to provide them “unfair” buffs later on).
I want more detail, what I’ve read so far about the differences seems like vague generalities to me.
I share the worry that you can put a lot of stuff in the game that the AI isn’t capable of handling. Building and pop placement on a planet map, while nice, definitely seems out of place to me in a game where your empire could eventually encompass hundreds of systems.
Since they have gone with this planetary economy minigame, I guess we just have to hope they’ve designed a minigame that their AI can competently play. Not just for the sake of the AI players, but for all the human players who don’t want to bother with it once their empire expands beyond a handful of systems.
It should be possible to design these things in ways that don’t trip up the AI. Like if the AI has problems doing long-term planning for buildings that haven’t even been unlocked yet, you let those new buildings be an optional upgrade for a basic building, with similar synergies, or otherwise make it easier for greedy optimization to succeed (or at least fail gracefully).
I don’t think you will get the details you want by asking such a high-level question.
However, I think we can infer a lot based on what they’ve said so far, and how things have worked in previous Paradox games. A Paradox-style war resolution mechanism is a big difference from most space 4X games. Victoria-style internal political modelling is a big departure. A Holy Roman Empire in space, where your nation is only part of a larger conglomerate is new. In short, Paradox has a history of putting a lot more into diplomacy and politics than most space 4x games, and there are signs that this will also be true of Stellaris.
I do like this post from one of the devs, in regards to “flavor” in the game:
I think that’s an important point. You get a lot of flavor from a WW2 game from the fact that you’re playing WW2 and you understand and are familiar with the setting. It’s one of the reasons that I never could get into the random map/country generator mod for HOI3. I thought it’d be cool for replayability since I really liked a lot of the game systems, but it lost so much be not having the historical context. Oh, look, this grey blob is invading and taking over this blue blob. Can you believe how good this purple blob is doing this game? It’s just not the same as seeing Italy win a (surprising) smashing victory in North Africa and setting the British reeling.
I am particularly interested in the AI because the PGS (“Paradox Grand Strategy”) games I’ve played (Vicky 2 and CK 2) were by their historical nature quite asymmetric and the AI was not nearly as important as a typical 4X. In CK2, I’d go so far as to say that “bad AI” improved the game because medieval kings acting like short-sighted jerks is part of the setting. Will this game be asymmetric in that style as well? I didn’t see it in any of the press stuff but it could be very interesting to have a space game like that.
I know there are several people asking for this, but all I’ve seen so far is that all the active races start with one system. There are fallen empires that have high tech toys but can no longer build or repair them, but as far as I can tell, no pre-existing large empires. It would be pretty neat.
Yeah, I’m a big fan and proponent of asymmetrical nations that exist in their previous games. I think it provides an interesting avenue to lower/raise difficulty without having to inflate the hell out of AI bonuses. It sounds like Stellaris is going for the most traditional “everyone starts out on even footing” approach, but admittedly that’s just my take on the statement that every race would start out on the cusp of FTL/interstellar travel.
Even if this is the case (and I agree with you all that it appears that it will be), an interesting “transition” mid-way through the game would be into the sort of game of thrones-y (I say never having read or watched that) politics of tiny, small, large, and massive powers and the chains of rule and subservience that connect them. If you’re faltering or stuck in a resource-poor area, vassalizing yourself to a larger power could easily be the “best” play long-term–especially if politics played out on multiple levels (e.g., stamping out rebellions and controlling spending within your personal sphere of influence; working with others in the alliance/empire you’re a part of and jockeying for position with your “Lord;” and finally the wars and alliances between large-scale states), enabling you to try to “take it from within” rather than conquer from without. All of this is informed by CK2, the only grand strategy from PDX I’ve spent any amount of time with, but it would fit well with their stated goals of distinct game-phases and advanced diplomatic/political options.
Hell, just the vengeful joy of being the “rebellious factions” causing trouble for an AI player a few orders of magnitude your larger (as compared to the similar, simulated rebellions taking place within your own holdings) could be pretty amusing in its own right. Pirate faction, anyone?
I suspect I’ve managed to wheel very far afield from the questions Brian was originally seeking (and the scope of the interview I suspect he’ll be “allowed” at this time in the game’s development/press cycle), so no worries about trying to condense that brain-dump into anything remotely askable, Mr. Rubin!
I’ve tried to whittle everything folks have wanted down to sentences. ;) I’ll be sending it out today.