Which 4X/grand strategy game has the best diplomacy

[SundiataWTF]

Is there a modern game that has a robust and semi-realistic diplomatic model? I'm looking for games that can be one on the diplomatic or economic battlefield rather than the literal battlefield. I don't mind combat, I just want the option to form grand alliances and trading partnerships that make it possible to win without world conquest.

[ashbery76]

This again.EU4,CK because it's tied into the system.

The rest all all the same imo and just copy the Civ formula ala boring as shit and nothing like real diplomacy.

SR2 is a senate.I like it but it needed more 1 on 1 diplomacy to go with it.

[Aleksei.Volchok]

You know, maybe it's just me being autistic but I never feel like strategy game personalities work.

So I like it in a game like Europa Universalis 4. There you have some ruler personality affecting things (more so in Rights of Man expansion) but it's not important compared to geopolitical situation. Most of it is clearly explained: they don't want alliance because they think we're weaker militarily and we're allied to their rival, here are numbers that allow rebalancing that stuff. But it's not just a predictable system, there are plenty of places where AI can surprise you like chosing rivals and expansion paths. This system is the best.

In all other 4X games AI looks more like a black box. It was OK in Civ4 I guess even though there it was relatively transparent. But even in otherwise great Endless Space 2 where AI literally says "I see you're weak right now and as I'm a sneaky guy I am now SNEAKY against you and will attack ASAP even though we have some agreements!" it still looks schizophrenic.

[pavig]

It's not so hard to program a tactical AI, but a strategic AI is very difficult. Unfortunately making the AI a but random humanises it so it doesn't play like a totally predictable robot, providing the illusion of strategy. Sadly the highly strategic thinking required of diplomacy is beyond the scope of the basic AI we have in most games.

[athelasloraiel]

Strategic AI without diplomacy that really excells - AI War(part 2 in development after successful kickstarter).

Ppl should really try to beat that AI (it doesnt cheat and isnt all seei g.)

Sadly, i have never met truly human immitating diplo AI. Would be nice to have one - that learns from mistakes, changes design, plays its diplo personality good enough.

Hope we see some revolution in that area while I live...

[Aleksei.Volchok]

Цитата допису pavig:

It's not so hard to program a tactical AI, but a strategic AI is very difficult. Unfortunately making the AI a but random humanises it so it doesn't play like a totally predictable robot, providing the illusion of strategy. Sadly the highly strategic thinking required of diplomacy is beyond the scope of the basic AI we have in most games.

I would argue the opposite.

In terms of strategy there's usually a lot of unknown and plans are long-term. In a game like Civilization it's hard to see AI's bad strategic decisions because you can rarely do something outright stupid. If AI declares war on you while being weaker he may be hoping for a dogpile, or he is desperate or you don't actually understand his strength.

But on tactical level mistakes are obvious and immediate. If AI leaves his archer undefended and it's not a trap then you know AI is dumb. And it's actually hard to program proper tactical AI. Most tactical games rely on AI being on defense and player being on timer - which usually means that AI can do precisely nothing and still win.

[pavig]

Цитата допису ilitarist:

...I would argue the opposite.

In terms of strategy there's usually a lot of unknown and plans are long-term. In a game like Civilization it's hard to see AI's bad strategic decisions because you can rarely do something outright stupid. If AI declares war on you while being weaker he may be hoping for a dogpile, or he is desperate or you don't actually understand his strength.

But on tactical level mistakes are obvious and immediate. If AI leaves his archer undefended and it's not a trap then you know AI is dumb. And it's actually hard to program proper tactical AI. Most tactical games rely on AI being on defense and player being on timer - which usually means that AI can do precisely nothing and still win.

I think what you're seeing there is the illusion of strategic intelligence. Even some solo board games can run what reads like a compelling strategic AI just using some rules and mecanics tied to a deck of cards or some other guided randomness. It's not intelligence per se as scripted motivation/direction. The problem is that it is strategy is a long term global/gesthalt understanding of game state and very difficult to codify.

Tactics though are a tractable problem for AI, in that they are predominantly local problems. This makes them solvable by solving the traditional "three moves ahead" problem of strategy games - ie. applying a few local guidelines, possibly simulating possible future results (which can be done using decision trees) and taking whatever move looks optimal at that point in time. This is pretty tricky to do, because it is actual artificial intelligence of sorts, but the various approaches have been around for decades and form part of the coursewok of most comp-sci degrees.

For the most part tactical AI that doesn't think ahead much works pretty well, which is why when it makes a mistake (such as not hedging for a bad die roll and leaving itself open) it is so bewildering to discover that it did something pretty stupid. Part of that may be bad ai programming, but part of it is the fact that tactical AI only needs a fairly shallow understanding of the game state to work most of the time, thus most are fairly shallow to avoid becoming bloated.

This is not the fault of the tac AI itself, which can't take on too much responsability as it may have have to deal not only with the current local game state, but be subordinate and tasked with implementing tactics towards the goals of the strategic AI. Given Strat-AI may have little understanding of game state and is just as likely to have picked a strategy at random from a few scripted alternatives it is no wonder the Tac-AI sometimes acts oddly.

But back to diplomacy in 4x games - most diplo AI operates on a tactical level, which is why so many have unsatisfying diplomacy play. The AI makes predominantly local decisions regarding the current game state, like "beat up on the weakest player", "ally with the player a bit more powerful than me" etc, without much considered regard for long term strategy.

The games which do diplomacy better tend to be the ones with the most complex hidden disposition systems, like crusader kings which throws a bunch of dice constantly to figure out who hates who, remembers forever, and lets seemingly intelligent behavior emerge out of that. Ironically there is no intelligence in the decision making of the AI in such systems - the intelligence is in our ability to tell stories to justify its mysterious, arbitrary but seemingly consistent behavior. ;)

[Mezmorki]

I also think that a lot of games don't pay enough attention to the end game and victory conditions, opting to slap something together the 11th hour before release. It isn't, therefore, any surprise that AI's don't really play to win since they probably aren't even programmed to work towards a long-term goal in the first place.

[Troy]

Цитата допису Mezmorki:

I also think that a lot of games don't pay enough attention to the end game and victory conditions, opting to slap something together the 11th hour before release. It isn't, therefore, any surprise that AI's don't really play to win since they probably aren't even programmed to work towards a long-term goal in the first place.

So true.