So the AI is getting a massive overhaul for v1.4. And I have to say, it's making a much bigger difference than we had originally expected.
To be honest, I was pretty skeptical that it was worthwhile to have engineering time spent on improving the way the AI manages its planets. Not to be a a lazy bum too much but I figured it a lot smarter to just let expert players play against the AI with bonuses. But as others have pointed out, the compounding nature of planetary improvements makes the bonuses we give too trivial.
Now, the AI is really quite good (better than most humans) at placing improvements in 1.4. It will never be as good as the best humans (traveling salesman AI problem) but it's close to an order of magnitutde better.
They're also working on improving the way the AI designs ships. Now this matters in a lot more ways than simply warfare. In GalCiv II, the AI took into account the ranges of various things to decide how much range techs it needed and how much life support it should have on ships. This is a non-trivial coding effort but needs to be done in order for the game to more challenging on larger map sizes.
Combined with the 1.31 AI update (the AI should be very noticeably better in 1.31 versus 1.3) I think that by the end of 1.4 the complains about GalCiv III's AI should be largely put to rest for most people. We shall see. The opt-in for it should be available in a few weeks.
Two people agreeing on the internet after an argument? What sorcery is this?
Guessing the mind control worked? Super villain confirmed!
Meh, it's happened before. That was on an AOL board in 1987, but it DID happen.
Anyway, I think we've covered micro to death now. On to Macsen's diplo point.
I don't think that diplo model is fundamentally broken. Sure, it's not very innovative, and it's presently very imbalanced and quite possibly bugged or incomplete, but it does the trick and it's actually a reasonably sophisticated example of the standard 4X diplo model once you poke around under the hood. In fact, it's somewhat more sophisticated than its behaviour in-game, which is part of why I suspect chunks of it might not be active or working yet. I'm not hugely excited by it, but it's not exactly disappointing that they're running with it either; the basic mechanical model is fine. Everything else is an AI issue.
I don't really think the BERT 'diplo currency' model is particularly original, or much fun either tbh. I've seen it on other games over the years, and really, it's very gamey and just as easy to exploit once you know what you're doing. I'd rather they stuck with the relative freedom of the existing model and looked at ways to make the AI more 3 dimensional.
Hi Naselus, glad you made up with eviator
I don't think that diplo model is fundamentally broken. Sure, it's not very innovative, and it's presently very imbalanced and quite possibly bugged or incomplete
The thing is can it ever be well balanced? Resources, Techs and ships all have different values depending on different circumstances within the game. (e.g. purchasing opponents Colony ship on t10 vs colony ship on t100) It's constantly changing. Setting an arbitrary value for each item doesn't work. Stardock highly likely aren't able to create an AI that can constantly assess value based on circumstance so whether that's broken fundamentally or not trades will always favor the player significantly.
Some players might enjoy getting one over on the AI in this manner but it's like taking candy off a baby, the whole system seems to exist to exploit AI weakness. If you try to make the most of your advantage in this area by mercilessly ripping off the AI as much as possible you end up spending half your game time in the diplomacy screen.
Whether BERTS original or not I don't know, but I don't recall a 4X game having zero! resource, tech or unit trades via diplomacy (the really exploity stuff removed). Most 4X games diplomacy systems are very similar so the smallest hint of any possible innovation in my eyes is a big thing. Yes the treaty trades can lead the player to exploit the AI still and not be a relatively fair deal to both parties but they are a hell of a lot more interesting than looking through 100's of generic techs and is probably much easier to balance so the player can't game the system so easily.The threat and respect levels which change via many conditions are a great way to show different factions personality and tie into the treaty system and the willingness of the AI to ally or declare war, it looks great. The games not even out yet so my opinion may change once I've got my hands on it but it looks vastly more interesting than Stardocks 1990's copy and paste clunky diplomacy system. From the previews I've seen BERTS peace deal warscore system needs some work, that's definitely not original (Poor unfinished copy of EUIV's peace deal system comes to mind)
I think a BERT like system could work with GC3 perhaps tie it in with Influence somehow. Anything would be better than the tired boring old broken formula we've been subjected to at the moment. Stardock may get knocked down for taking risks that fail, but they deserve to get knocked down even more if they don't take any.
lol
Don't knock old -- for tech trading, I personally still think the *original* Master of Orion got it right (not the crappy MoOII, the *original*) -- when you want to trade, the AI provides a small list of techs they're willing to trade, and when you choose one they provide a small list of techs they'll take for what you want; they never offer their newest and best techs, and if you aren't willing to trade one of the techs they ask for (e.g., as a rule I never trade an Improved Robotics Control tech that's not at least two levels behind what I have) you're out of luck.
For that matter, I like the way trade agreements worked in MoO as well -- they cost both parties money at first, and it wasn't for quite a while that they made money, so both parties had an incentive not to go to war immediately after signing a trade agreement; I don't think I've seen that anywhere else.
Recent games haven't done as well when it comes to diplomacy perhaps because the diplomacy is less abstracted; I suspect that diplomacy is one of those cases where abstraction is a good thing, and while a more detailed implementation _sounds_ good it never quite works in practice because a human has too much of an advantage over the AI (espionage is probably another -- it's also another case where the MoO implementation is possibly my favorite).
The thing is can it ever be well balanced?
Yes, I think it can.
As I say, there's nothing mechanically wrong with the diplo interface. When two humans seek to barter together, it does the trick. Diplomacy need to offer that kind of freedom. The problem is an AI one. And I'm not really speaking specifically about GC3 atm; this is true of most 4xes, which have near-universally adopted the two-menu approach simply because it's the best choice mechanically. It's obvious what everything is, you can ask for most things and trade most things, and it allows you to engage in reasonably sophisticated diplomatic behaviour. Mechanically, it's a very sound model; much more so than any of the alternatives. Adopting something less would be a step backwards, retreating from the challenge rather than overcoming it.
The problem is, of course, the AI. I've never been one to favour making the game easier so the AI can play it when you could be making the AI better. I don't want to play a game that's dumbed down to the point bad AI is good at it. I want to play a good game that the AI needs to be written to a high standard for. And, poking around under the hood, we find that GC3's diplo AI genuinely has been, to some extent.
I'm not convinced that the diplo AI is actually entirely functional - several of the diplo value files seem not to do anything, including the one that defines the AI's acceptance limits - and there's plenty of additional modifiers I'd add to it's valuation calcs, but they're actually very good. It takes account of a really wide array of things, though the values it takes for them aren't really balanced properly yet. It has relations memory, 60+ relations modifiers (which have different impacts for different personalities), and takes up about half the AI defs files. It even applies modifiers to it's valuations in some cases (though it could do with many more).
The big change I'd make isn't a re-work of the whole module; diplo isn't really the issue. I'd just add a couple more relations 'types' to make the AI more three dimensional. The real problem is that it doesn't differentiate between liking someone because you're trading with them and liking them because you're scared of them. The AI is easy to manipulate because it doesn't have multiple motives, like players do; it can't be ambiguous. If we had, rather than just 20 possible 'relations' scores, instead had 20 possible 'love/hate' scores, 20 'fear/condescend' scores, and 20 'trust/distrust' scores, all being impacted different amounts by different actions, then the AI could do things for different reasons. And show them to the player.
So, say we take alliances, for example. Presently, the AI will ally with you once you have relations of over X. It doesn't care how you got to those relations. You could have trade treaties. You might have killed pirates near them recently. Maybe you have open borders. Maybe you inverted the military score modifier and it's scared of you. Who cares? It just likes you.
With a multi-dimensional AI approach, you instead can get an alliance if it fears you OR if it likes you. Either scare the crap out of them, or trade with them and play nice. But that's not all. You can make it more likely to break alliances for multiple different reasons, too. If it's only helping you because it fears you, maybe it will break the alliances more easily. If it likes you AND fears you, maybe it'll still refuse the alliance if it feels you're untrustworthy. Maybe it hates you SO MUCH that it won't buddy up, even though it's terrified of you.
This'd give you the kind of unpredictable, human-like AI that Civ 5 attempted to get on the cheap by just not showing you the active relations modifiers. And it'd let you show the player, so he's not mystified as to why the hell Gandhi just nuked Calais when they'd been getting along famously. He can see the AI's motives, but it's hard to predict them without memorizing 20^3 different relations combinations.
Agreed a two menu approach is fine.
Yes, but they are "bartering" that isn't really what's going on with human to AI. An AI can't assess things as far as I'm aware.
It's the "bartering" part of diplomacy I have the real problem with, relations I'll leave for another day.
The thing is I think Stardock have given themselves an impossible or near impossible task. Yes the AI is the problem if they could get it to think like a human and do trades like like a human then it wouldn't be so easily exploited but that's basically impossible to do.
The way I assume they go about it is to assign values to items and if the player meets the items value then a trade can be made. Attaining even a small amount of balance between 100's of items would take an enormous amount of time I imagine. It's no wonder Stardock basically haven't even tried yet that task must be daunting. And like I mentioned before the value of things changes based on circumstance also, which would take even more work on their part to achieve any semblance of balance. I don't think it's a case of dumbing down it's a case of being realistic, it's not really possible to achieve.
But this is the key thing with GC3 Diplomacy trade system, it's not even even fun! Even if it they were somehow able to achieve some level of balance where a competent player doesn't believe they have a massive advantage when trading it would still be terrible. I simply don't think it's worth the work and effort they would have to put in. Scrawling through lists to trade generic techs resources and ships is boring as **** as far as I'm concerned.
And the thing that adds to the levels of micromanagement and utter tedium of the system is the SCALE of Galactic Civilizations 3. It's a system not designed for a game that can have such a huge amount of opponents. Older games from 90's early 2000's didn't have that, it was a lot simpler more manageable, it was also newer but still not especially engaging. GC3 has 100's of items, generic techs, resources, ships and up to a hundred opponents with poor tooltips and no time saving measures to help the player trade whatsoever. You can't enter one screen and immediately see what everyone has you have to enter each diplomacy screen individually. The system just isn't fit for purpose for a game on this scale.
Yes. Why you are at "Friendly" at the start needs to go away. You should have to work hard for tech trading benefits etc. That is the main fault with the current system, not AI or whatever.
Yeah, that's been a major bugbear for me too. it would be good if the dialogue option you have when you first meet new races had some impact on diplomacy, including an always negative of you haven't researched universal translator.
Diplomacy is the number one thing that makes me bored in this game as it stands now. Whilst the brilliant new AI in 1.4 will make it more challenging, I can't help but feel that this will make which races you play against redundant and lack personality. Nasaleus, your suggestion of multi-faceted diplomacy decisions would be welcome, as the main reason I began playing this game was because of how the AI in previous GC games was unpredictable and you could really feel the differences in races that you were playing against. GC III as it stands now feels like im playing against AI bots after turn 150.
Better feedback with diplomatic modifiers and a multi-faceted AI decision making process (re: trade and relations) would make this game really stand out above its competitors.
Sorry for double post, but also allowing the player to make requests (gifts etc) and threats in the same way the AI currently does would also increase that immersion with the game.
The 'get off my lawn' stuff for diplomacy IS going to come. It will go a long way towards helping with immersive-ness.
I'd like to see the AI 'personality' reflected in what it will trade and what it expects. So for example a money-loving race might highly value commerce techs yet be willing to cheaply trade away trading techs. Militaristic races should want a lot for any military technologies but not care much for anything else, and so forth.
That does open up the 'buy it cheaply from X and trade it to Y who wants it a lot' gambit, but... that's what commerce is. And the AI could take advantage as easily as the human player.
Either that or toggle tech brokering if that idea annoys you.
I would like to see a more diverse AI that plays in line with the personality traits the race has (such as 'aggressive', 'greedy' etc.) I know it does this to an extent, but I'd like to see more of it even if it's more personality choices to pick from, such as 'Isolationist'. This AI type would expand slightly slower, but avoids getting into wars or trading as often, however players without open borders in their ZOC will suffer a large negative diplo. modifier, and are less likely to enter into treaties overall.
Not only would this liven up the AI behaviour and make them more interesting (as opposed to a race for military power and then war) but will also have the roll-on effect of making custom races feel more alive also, as these personality types could essentially be modular AI behaviours that players can choose in order to determine how their race acts.
^^^^ This would work for me. I know Stardock has a LOT on its plate with releases and its engineers working on projects. A major diplomacy update would be nice. I would love to see the Ai's we have (the core ones at least) be given something like the above in its trade valuations in diplomacy.
If its moddable or can be I would take a shot at it but I do know such weights are either not in or not working....
Ok there is much more but thats the idea. Also this idea coupled WITH disabling tech BROKERING would go along way to getting the ai to 3-way deals with you to get what it wants.
I'd also like to see it approach you to say, "hey if you research intensive farms and give it to us we will ally with you and be your favored trade partner'.....little NON-warefare diplomo perks like that would also go a long way to make the Galaxy come alive.
There are many great features available to you once you register, including:
Sign in or Create Account
