Fall patch discussion thread

[Siesta Guru]

Out of curiosity, how did the AI work in Civ5 exactly?

It had mostly unique implementations inside the dll for all of the different decision making areas, no on-size fits all solutions. (which isn't without its flaws either)

The war-operations thing was mostly just used to get its units to the front. It formed 'formations' of units, then sent them over to the war zone. And once they got there, they would mostly lose their cooperative nature and would basically be a bunch of individual units doing their own thing (with some semi-cooperated focus damage on enemy units). It would originally just loop through all of the units once, and pick one action out of some set of possibilities. This was why they wouldn't be able to move and shoot, because that's basically two actions.
There was little in terms of 'thought' either, about what the best way of handling a situation was. But since it was rather simple, it would at least get the units into the action doing stuff, even if it was stupid stuff.

Settler escorts were coded completely seperately afaik, and it would just have a settler and another unit move towards each other and would then march to the intended destination together. That simple system worked remarkably well compared to what we have now.

Fascinating stuff. Would this mean that even if Firaxis releases the code, there is nothing much modders can do?

Well, with code access, we could always rip the entire existing system up and make our own from scratch. But that kind of thing tends to be a bit too much of a lengthy and risky endeavor, so I'm not sure if anyone will. An additional option would be to heavily expand on the behaviortree system and add our own nodes and change the behaviors of current ones. That could possibly increase AI competence meaningfully, but to what extend I don't know yet. It depends a lot on how it's coded internally.

 

[rezaf]

I simply don't understand why this is the case, though. Surely, Firaxis must understand that AI is one of the most important things in a game like CIV, and that is has to work properly?

Is it?
Civ5 was the first civ to allow what amounts to casual players to beat it on Deity - thanks to 1UPT and the AIs inability to cope with that.
Yet the game got glowing reviews, endless props from many many players (who sung high praises about the very thing that the AI struggled most with) and sold like hotcakes.

And now Civ6 is out, with an AI even less capable of putting up a challenge, yet - once again - glowing reviews, endless praise and excellent sales. All this with an AI that does NOT work properly.

So, why should anyone at Firaxis care?

 

[Siesta Guru]

I simply don't understand why this is the case, though. Surely, Firaxis must understand that AI is one of the most important things in a game like CIV, and that is has to work properly? Quite frankly, it sounds like too little resources/time was spent developing the AI, and that simply isn't good enough. I hope there is a push to completely revamp the AI within Firaxis right now. Anything else would be disappointing.

Yeah I'm pretty sure you're hitting the nail on the head there. Not enough resources were spend on it. There's probably a dev out there reading about the AI comments, agreeing, but without having been granted the time to work on it. Programmer time tends to be a highly valued thing in these organizations, as they often work with small teams of programmers (theyre bloody expensive). And they probably needed to use that time to fix even more game-breaking stuff that just had to be ready before launch.
My bet is they only had enough programmer time to give it one shot at developing something appropriate and couldnt afford the time for a full do-over when it turned out to be lacking. And of course you don't want the programmer to be mulling through xml files, hand-tweaking variables untill its as close to good as possible, you're going to give him just enough to put a random nice-sounding number on paper and run with it. Most of the extremely important AI values specified in the xml are "1" or "0.5"

 

[LDiCesare]

Is it?
Civ5 was the first civ to allow what amounts to casual players to beat it on Deity - thanks to 1UPT and the AIs inability to cope with that.
Yet the game got glowing reviews, endless props from many many players (who sung high praises about the very thing that the AI struggled most with) and sold like hotcakes.

And now Civ6 is out, with an AI even less capable of putting up a challenge, yet - once again - glowing reviews, endless praise and excellent sales. All this with an AI that does NOT work properly.

So, why should anyone at Firaxis care?

Ye, I think that's the point. Graphics matter. AI does not. I'm telling everyone not to buy the game till it's patched, and since my colleague and I work onAI, it's easy to convince him, but for most players? I know I won't buy expansions or DLC if they don't fix the AI, but does that have any significant impact on Firaxis?

 

[Sic]

Not until reviewers start looking at AI critically.

There is a tendency amongst reviewers today to create little stories around strange AI behaviour, and especially after rougelikes made a comeback (RPS, I'm looking at you). People have gotten so hungry for games to be story generators that they've grown increasingly blind to actual problems in gameplay. "As long as it's fun" is an OK adage to play games by, but everyone needs to be able to see that AI that repeatedly is not able to do simple tasks is not something that can be attributed to some deep level of "story algorithms" aimed at making the computer seem human or unique.

I think in the case of CIV6 this tendency was furthered by having leaders that oozed personality, and was promised to have very distinct behaviour. There were quite a bit of "well, they probably had their reasons" arguments going around.

Now, after potentially playing a multitude of rounds of CIV6, I think it would be fair to amend reviews and ratings, given that a lot of them were made on false premises. If there is one language developers and publishers understand, it's the language of Metacritic.

 

[dexters]

The first patch of fixes are promising and even then it feels more like a glorified hot fix , as I noted upthread I've noted improvements already. But always room to improve.

For some the AI will never be good enough, not that it was ever good in any of the games. The only thing 1UPT doesn't allow is for the AI's massive production advantages to kick in by overwhelming the player.

I sort of understand the 'Diety is totally possible with 1UPT' by 'casual' players arguments, but those wins tend to be kind of cheesy with baked s 'rigged' starts anyways, but even that is old, people have been beating diety or archipelago maps with hand picked civs for ages.

 

[oedali]

This is so interesting. I just could not fathom how Firaxis failed to get the AI to escort its settlers, given that the AI has been easily able to do this in past Civ games and that Firaxis even created a fancy new escort feature for Civ VI. But your explanation makes sense.

I got about 10 games out of Civ VI, which is a pretty fair amount. But I think the AI is so bad that I can't play it much more. There are so many cool new features and so much potential. But the combat part of the game isn't fun if the AI mills its units around and doesn't attack, and the expansion part of the game isn't fun if half the AIs can't get past three tiny cities on Deity. The fall patch didn't really change these things. So probably back to Civ IV for a while for me.

I feel the exact same way. Have been playing Civ 6 since it came out but enough is enough - AI sucks so bad especially in combat/operations that it takes the fun out of playing the game, even though Civ 6 has so much potential and there are lot of things I like about it. I am going back to Civ 5 community / VP mod where they AI is very good at fighting & everything else. I don't want to play every game knowing there's a >90% chance I can win. I also still cannot comprehend why the devs didn't leverage the Civ 5 VP AI code for at least the tactical/operations stuff and instead decided to go back to square one with this "behavior tree" system that obviously does not work yet.

 

[Sascha77]

Maybe they were trying to be too realistic in their game. I know in the real world Europeans went up against "civilizations" such as Native Americans who probably only had archery level tech, but it doesn't work well in a Civ game. I wouldn't mind one civ being backwards and behind, but not all of them. Especially if they aren't isolated and have contact with the other civs in the game. Technology and resources do spread beyond borders. Even if because people wanted to make money, especially if people wanted to make money.

Well.. like you say: Under certain circumstances (like a civ/civs developing in an isolated part of the map), this can be viewed as realistic. The Aztecs were a highly developed and pretty advanced culture - their only problem was that they were basically fighting Spanish longswordmen/Musketmen with Stone Age weaponry.

Personally, I have more of a problem with mid-/late game barbarians popping up everywhere and wielding top-of-the-line weapons/units. At least in Civ IV, these guys could run their own cities (not just "camps"), so it made sense that they'd be able to produce weapons more modern than clubs and wooden spears.

S.

 

[Emi999]

Well behavior trees are used a lot these days (which doesn't automatically mean they are the best solution for all types of games). This is a good start if you want to know more: http://www.gamasutra.com/blogs/ChrisSimpson/20140717/221339/Behavior_trees_for_AI_How_they_work.php

But yeah, they DO have problems when it comes to reacting to events and when trying to coordinate multiple actors together. But they also allow tons of moddability (particulary if you can code new nodes).

 

[Agent327]

AI with neutral relationships should accept delegations barring exceptional circumstances.

Not sure if it's been mentioned, but this simply does not work. In my experience the AI will simply refuse delegations (yet ask for delegations to be admitted) if you don't ask immediately after meeting a civ. It will accept delegations with declared friendship, but not if this occurs after a war.

Practically this means that in any game you will have only delegations with civs that stay friendly (low chance of that happening), and permanent residencies will only be accepted by one or two civs (yet all will ask for such).

This is simply not realistic diplomacy.

 

[CivLuvah]

I saw there is this icon on tech and civic trees in my game post-patch and when I hover it it says "recommended". Is this part of the patch?

 

[Stilgar08]

I saw there is this icon on tech and civic trees in my game post-patch and when I hover it it says "recommended". Is this part of the patch?

Spoiler :

yep. To help you along pursuing your wanted vicotry type or current priority... Not too bad IMO...

 

[ThERat]

how does the game know what victory type I want? It never asked me.

 

[Sic]

how does the game know what victory type I want? It never asked me.

Yeah, I don't like this at all. I hope we can turn it off at some point.

 

[Cymsdale]

Yeah, I don't like this at all. I hope we can turn it off at some point.

Recommendations have been in pretty much every civ game, no?

Just ignore it, not that big a deal.

 

[rezaf]

This is simply not realistic diplomacy.

Lol.

You DO remember we're talking about the game in which an AI civ declares war on you, you fight them back, after losing a couple of units, they offer you peace on favorable terms.
Some years later, they declare war again, with the same outcome.
And then they do it again.
Afterwards, when peace has been declared for the third time, after the third war they have stared, they denounce you for being a warmonger.

There's no diplomatic victory because meaningful diplomacy isn't implemented.
On one hand, I'd hope they'll fix this in an expansion. But on the other hand, unless they do some serious fixing in free patches first, I'm not sure if I'm inclined to purchase said expansion. I'll probably get it heavily discounted at some sale, but full price?
This'll take some convincing updates.

 

[Agent327]

Why are you repeating my comment? 'Civ 6 doesn't have realistic diplomacy.'

Oh, eh... LOL

 

[Siesta Guru]

Well behavior trees are used a lot these days (which doesn't automatically mean they are the best solution for all types of games). This is a good start if you want to know more: http://www.gamasutra.com/blogs/ChrisSimpson/20140717/221339/Behavior_trees_for_AI_How_they_work.php

But yeah, they DO have problems when it comes to reacting to events and when trying to coordinate multiple actors together. But they also allow tons of moddability (particulary if you can code new nodes).

In theory the concept can kind of work if a lot of effort is put into it. However, in civ 6, you cannot code nodes and the nodes present are incredibly limited. There's almost no conditionals, the few variables that can be set are arcane in usage, there's no gotos, and the way it selects which of its nodes are childnodes may be buggy (or I don't understand it properly yet). Worst though, most of the leafnodes are on a very high level, without any control over the finer details. For instance, the only way to interact with enemy units at all is to use an 'attack units' node. You cannot specify what units it should attack, which units in its group this should apply to, etc.
Like in theory, it could be useful and very moddable. But in reality, the tree that makes it choose what improvements to build is a bit of fluff around a node that says 'choose improvements'. It would require a signfiicant overhaul to even get this to the point where it can be used for okayish smart decision making, and would never even approach something like a neural net implementation.

 

[rezaf]

Finished up my first full post-patch game and I think I'm gonna shelve the game for the time being.
There's a solid game somewhere therein, but it's disheartening to see how much work is still left to do.
Even leaving the AI aside, the UI still leaves a lot to be desired, pacing and balancing are totally off, diplomacy is broken, religion is a Moderator Action: <snip> and too much of the game resembles a chore more than entertainment.

In this game, I've not seen many carpets of doom,but not many coordinated forces either. Usually, the AI declared war on me without forces on the ready and then sent in a tickle of troops whom I could defeat piecemeal. I don't really know whether this is worse or better than my first games, where the AI would phone in it's "surprise war" a dozen turns before the fact by sending in so many troops it fell over it's own feet.
At least the AI DID build many corps and armies - though too often of long outclassed troop types.
I also actually encountered the scenario I was kidding about earlier where I was trying to besiege a city but could not get in range because two civs neutral to the conflict were duking it out with their apostles.

It's also kinda fishy that AIs vastly outclassed militarily are acting the way they are. I hated the idea introduced in Civ5 that an AI should be "playing to win" in the same way a player is, but at least I could understand the reasoning.
Here, I'm getting demands (they didn't even bother to make an extra exchange for this, the AI just suggest "trades" offering nothing in return) from Civs I'm way superior to.
Third world powers with medieval forces declare war on me, facing off their pikemen against my modern tanks, helicopters and rocket artillery.
I get trade offers for half a dozen luxury goods in exchange for one and 2 gold per turn or something - why can't the AI suggest (or agree to) something that's mutually beneficial?
If I NEVER declared a single war, why must everyone denounce me as a warmonger?
I'd take the Civ1 AI over this mess any day.

Due to the AI's inability to properly wage war, I'd been holding back on going for a domination victory (the AI is so bad this feels like cheating) but went for a science victory. When I first tried this, it was my very first game, and I didn't know a thing about planning cities to maximise factory overlap or stuff like that, so all the space projects took in excess of 50 turns and I ended up accidentally winning a culture victory instead while building the first mars project.
This time around, I was a century earlier and HAD planned a cluster of cities, complete with the Ruhr Valley and whatnot, yet in the best city (the one with the Ruhr Valley), each of those projects still took 28'ish turns, 35ish in the other cities of the cluster.
Since you cannot rush these projects, I feel this is far, FAR too long. Not to mention it's kinda counter intuitive that you now need to be the best civ at producing things when you were going for a science win.
I was still not finished with the moon landing when I noticed a few nudges should get me over the culture victory threshold, so I went for that again to avoid having to brainlessly skip turn after turn a hundred or so times.

Ah well, like I wrote, I think I'll shelve Civ6 for the time being.
How do you guys deal with those issues? Or aren't they issues for you at all?

Moderator Action: Please comply with our rules regarding use of inappropriate language.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889

Edit: Had used some word involving cluster and was asked to edit, but I think the moderator action thing gives the (omitted) description just the right kind of gravity. Thanks Browd!

 

[BuzWeaverCiv]

Does anyone know why we don't have a Population Notification?