Is AI really brain dead?

[fatgordy]

I know it's been said before, but I wanted to share another instance of the AI just not even trying. Immortal difficulty with all victory conditions open. Turn 144 and the AI still only has 2 cities while there are clearly some opportunities to expand. I haven't been messing with them and I could easily send settlers to take the open spaces but I wanted to see if they would expand. They just have that one settler sitting there in their capitol and it won't move. Why? They could take over that City State and expand up to the north. They could build up an Army and try to attack me. I only have archers after all and have felt no need to build any other military units. Why did they just give up? I have no desire to continue to play on this map except that I want the dual map achievement.

AI not even trying.

• fatgordy
• Nov 4, 2016

 

[ThunderLizard2]

They should not have released the game with the AI in the state it's in. It's not just a few bugs but it's totally busted. It's alpha level at best not even beta testing ready. Your example is a great one as it shows the AI can't handle basic management of cities and expansion which should be fairly straightforward. Combat AI is so much more complicated and the AI is just broken on all levels.

 

[Horizons]

The AI is good at spamming religion. I had a surprise loss due to religion once lol. But otherwise I was facing archers and chariots up against my tanks and planes (at Prince)

 

[Edgehopper]

Similar thing in my game--America must have been stuck on 3 cities (It goes Washington, New York, Philadelphia, Boston, right?) I'm playing as Montezuma on Emperor, and Teddy is being ultra-aggressive. The map is set up so that there's a mountain range blocking most everything to my north, oceans to the east and west, and Mohenjo-Daro set up on my NW corner. Japan is to the north on the other side of the mountain range, Teddy's to my NE. There's enough room on the land by me marked off by natural barriers for 4 solid cities, which I settle. Teddy forward settles Philadelphia on the little bit of space to my NW between the mountains and the sea; not a great place for a productive city, but effective for cutting me off. He then proceeds to fight with me for most of the medieval era, and the natural barriers make it tough more me to take Philadelphia.

While he's doing that, I see an American settler getting escorted around the sea near Philadelphia by a heavy chariot, both wandering aimlessly. Eventually, I get the units in position to capture Philadelphia and expand NW. And once I get past Philadelphia, about 10 hexes away and right on America's border, I find the Dead Sea sitting next to salt, iron, niter, and other great tiles--unoccupied! Of course I promptly settle it...at which point, after Teddy has agreed to peace, he sends a settler to found Boston on a useless sliver of land to my south with no redeeming value other than two wheat tiles??

 

[Shorlin]

This just happened in a standard immortal game playing as Tomyris.

I am in the process of wiping out Brazil. There is only Brazil and India left on my continent. India is also in an unrelated war with Brazil.

I conquer Brazil's second to last city, and the next turn, Gandhi sues for peace with Brazil and GIFTS HIM 2 OF HIS CITIES!

I mean I know Gandhi was known for his selfless spirit but come on...

 

[skyclad]

Wow, Gandhi keeps being an inspiration in selflessness! This belongs in the funny screenshots thread

 

[Lord Shadow]

More than one year ago, Soren Johnson (the lead designer of Civ IV) was asked his thoughts about 1UPT in a Reddit AMA. His reply was diplomatic:

I am very glad they tried it as it was certainly on my short-list of ideas for Civ that hadn't been done yet. Obviously, 1UPT creates some serious AI challenges, so I think your opinion about the mechanic is largely colored by how important a competitive AI is to you. (A lot of Civ players just want to walk their way through history and don't even like fighting wars.) I will say that I am very curious about what happens to the mechanic in future iterations of the series.

I left him the following reply -

I think Civ5 shows why you cannot have military strategy and military tactics in the same game. To see why they are incompatible consider a case where your civ is invaded by a much larger AI. If a small group of units can hold off an arbitrarily large enemy using the right maneuevers and tactics (Civ5) then there's little point to all the infrastructural effort the AI put in to build his large army. Conversely, if 30 units always beat 10 units (Civ3&4), then there's not much point to modeling all the maneuvering on the battlefield because even the most gifted tactician can't change the outcome.

In a game where strategy reliably beats tactics, that necessarily means tactics reliably loses to strategy.

It comes down to asking, which is more relevant to civ as a game genre, strategy or tactics? And I think the answer is clearly strategy. The military subsystem interacts with the rest of the game in strategic terms. It's all about opportunity costs, building a unit means not building a temple or lab, researching a military tech means not researching an infrastructure technology and so on. The stack of doom is a natural phenomenon and the only "problem" is it took too many clicks to manage compared to the gameplay relevance of what was in each stack. The answer was to simply make stacks an explicit gameplay element - "armies" would absorb "units" as you built them and you would only have to think about 5-10 "armies" on your map at a time. I think Civ5 went in the wrong direction. Tactics dominate and the strategic tradeoffs become less relevant.

Civilization series is ultimately a game about producing things... units, buildings, cities, technologies. The game is about snowballing your capacity to produce and prosper. A too-tactical focus takes away the whole point of the game. Playing battlefield general is fun, but when I play a civ game I want my civ to rise or fall because of the broad opportunity tradeoffs I made that define my civ's "character." I don't want to be able to battlefield-micromanage my way out of a war that I should lose because I didn't do like Stalin and order tanks built 10 years ago.

Soren replied I think that was a very good analysis.

The moment I heard Civ 6 was going to be 1UPT I knew the launch would be a trainwreck. But they were simply never gonna do it any other way. Civ 5 sold over 6 million copies and Civ6 has already sold 1 million at launch. Firaxis is chasing the money, people want a tactical game.

But ever since IV the Civ series has been all about adding complex new subsystems to the game without any regard for whether the AI can handle them.

When you boil it way, way down the AI interacts with the player in 2 ways:

1. Fighting the player in wars
2. Outracing the player to a nonmilitary victory condition, which is just a way of saying out-producing the player

The traditional Civ solution for making the AI competitive on the 2nd front has simply been to let it cheat its way to massive productivity. That massive productivity was then also the main thing keeping the AI competitive in wars, as its usual strategy was an overwhelming stack or armada of units. This was true in all the Civ games including the "good ones" like II and IV.

The problem is that in V and VI, productivity doesn't win wars by itself. Because the focus is tactical not strategic. But look at this thread, players don't want to win tactically, they consider it a BUG when the AI "throws its units away" in matchups that favor the player (due to positioning, terrain, unit matchup etc).

Yours is a very good analysis indeed. The question is, is the more deterministic "30 units always beat 10 units" actually more fun? 1UPT provides more depth and choices to the player, and wars have a meaningful progression beyond the cities' production line. And it allows smaller numbers of units to defeat larger armies through superior intelligence (always the main advantage a human has over the AI), but by design I doubt the smaller combatant is meant to do so indefinitely. In every 1UPT game, after all, no smaller army is meant to be invincible.

The AI has always thrown its units away. It was considerably more suicidal in Civ4 and before. But some of that reckless nature needs to be recovered so that the AI doesn't avoid attacking and causing damage for fear of losing units of its own. That's how a productive edge would be competitive: if the AI were more decisive and sent a reasonably continuous stream of units at the player's defenses, that smaller army of theirs would be gradually depleted and ultimately overwhelmed.

But in the end, as your analysis implies, opinions on the subject are also coloured by how much you desire Civ to be a production and numbers game. 1UPT heightens the wargame side, which is to some extent at odds with the production side. But if the 1UPT road had not been taken, I'm not sure how Civ's wargame side could've evolved beyond smashing stacks of doom against each other: if you keep warfare strictly strategic, its complexity has a (pretty low) ceiling, one Civ4 had already reached, from my point of view.

 

[Nixalo]

But in the end, as your analysis implies, opinions on the subject are also coloured by how much you desire Civ to be a production and numbers game. 1UPT heightens the wargame side, which is to some extent at odds with the production side. But if the 1UPT road had not been taken, I'm not sure how Civ's wargame side could've evolved beyond smashing stacks of doom against each other: if you keep warfare strictly strategic, its complexity has a (pretty low) ceiling, one Civ4 had already reached, from my point of view.

This is one of the issues with Civ. Everyone has different desires of how the game is played.

1. The Human's intelligent production empire plans vs the AI's difficulty based production bonuses
2. The Human's intelligent strategic picks of techs and civics vs the AI's free techs and civics
3. The Human's tactical use of era appropriate units and diplomacy vs the AI's hordes of units fueled by bonues
   
4. The Human's tactical use of era appropriate units vs the AI's tactical use of units

If you want a game where the humans must hold of the AL until they catch up to the AI and autowin, Stacked units is your jam. Because it's just about numbers and catching up via better management before you get steamroll. This doesn't win some fans as it gives you a very limited time to catch up or you lose.

If you want a game where the humans must react better to the changing eras, resources, terrain, and borders, you want 1UPT. This however make the game most wargame-like and yo either need better AI, a good personality system, or well crafted maps.

 

[Haggbart]

My main concern with the AI is that the game is too easy om immortal/deity. I do think this can be fixed if it would make a priority of building the right districts and buildings, upgrade its units (or even fix the upgrade paths), use trade routes and expand more.

I started on prince and was DOWed early on every difficulty level up to immortal, but not on deity. And when you get up your trade routes and factories you can you steamroll it whichever way you want, the difficulty level doesn't make much of a difference.

So my main concern now is running out of a challenge, which is allready happening.

 

[Stilgar08]

No adjustments needed

another "The AI is pretty bad"-thread? How refreshing!

Agreed, the AI should be more agressive and if it isn't it should withdraw instead of waiting to get slaughtered. It should be possible to pose a real threat. It doesn't at the moment (or at least it seldom does...). It should get improved...
Will it? Yeah! Will it be tomorrow? Nah!
Will it be by the devs or the community? I don't know - A bit by the devs. A bit later and more thoroughly by the community, that's my guess.

Does the game suck because of the AI issues? Maybe for you but not for me, although I agree it could be better... Personally I'm quite sure they (Firaxis) or the modders (not me - I'm too thick for that) will help us out... Next time wait before getting the game and you'll be fine!

P.S.: Thank you for the opportunity - I will now happily copy/paste this, maybe do a few adjustments and post it elsewhere, where they deal with the same topic...

 

[kampori]

America must have been stuck on 3 cities (It goes Washington, New York, Philadelphia, Boston, right?)

Names are now randomised after the capital.

 

[TurboJ]

Only played the game for two weeks and I'm at a point where there's no challenge left anymore. I feel it's time to maybe but Civ VI on hold and wait for patches / community improvements on the AI. There's no point playing when losing is never an option.

 

[TurboJ]

AI is (usually) not something you can do halfheartedly; it takes effort to make it work well. Every fix to one stupid behavior will create a different stupid behavior. And might not even fix the original stupid behavior.

But it's the same company that did Civ V; how is it possible the new AI is vastly inferior in almost every respect to what they had before? Playing on deity now feels like playing on prince then.

 

[Sascha77]

The AI is good at spamming religion. I had a surprise loss due to religion once lol. But otherwise I was facing archers and chariots up against my tanks and planes (at Prince)

The thing is that they *can* be pretty close in tech, so the AI clearly knows how to generate science/culture. In my Japan-game, Victoria had ironclads around 1000 AD (though she obviously beelined to steam power... perhaps she got a good combo of random boosts via a GS?).

However: They don't seem to "understand" the importance of having a modern army. In my conquest game as Trajan, England was still on a carpet of doom of archers when I invaded them with battleships around turn 280 on epic - and she did have much better tech than archery at that point.

Another critical error they commit is to completely ignore production-buildings. I've seen this in numerous games now where the AI simply will not build industrial zones. It's actually easy to check, since usually you'll be the only one competing for Great Engineers and there won't be many/any foreign trade-routes that give you production. In my (pretty ineptly played) science-victory game as Japan, only one AI civ (France) eventually started building these districts when I was already constructing my spaceports. I mean.. I know they get substantial buffs to their production-rate, but even those can't be enough to outweigh the disadvantage of not having workshops, factories or power-plants.

S.

 

[fatgordy]

Update on turn 200. AI still hasn't founded a third city; that settler is still sitting in her capital. Large mass of units now sitting outside of my borders (and in some cases, in my borders since we have open borders). About 50 turns ago AI started spamming apostles like crazy into my borders and converted two of my cities before I was able to build enough religious districts to compete. But now I'm able to spam inquisitors whenever I need them and use the in friendly border bonus to fight off the AI. Researching Rocketry now on the way to my science victory.

I did notice that the AI has been focusing a lot of resources on building wonders too with 5 combined wonders in her two cities. So I'm going to have to keep an eye on the culture victory, but I think I have that one covered too as I'm even in tourism and I'm closer to getting computers.

Turn 200 Update

• fatgordy
• Nov 5, 2016

 

[Heikki Mustonen]

AI has huge problems in finishing the game unless they get religious victory by accident. I have only had time to play four full games. I started with king and got easy science victory with India. Then moved to emperor and won easy domination with Rome. Third game was on immortal and I played Russia and won culture victory. In all those three games there was no doubt in any point that I might lose the game.

Last game so far was deity difficulty with Germany. I got quite poor start, got forward settled with two AIs (Norway and Arabs), got attacked by Norway and only got 3 cities up fast. The attack AI did was a joke. It had enough units to kick me out of the game but it basically just run around in circles until I got walls up and few more archers up and killed his units. He then made peace with me and gave me all his luxuries and money.

A bit later I managed to get to five cities with one more settle and when I captured one cite state. AI was also very aggressive against city states and in the end only three survived to the end. At this point the game looked very bad. I did not get religion in the game and Scythias religion had already over half of the world converted. Also Kongo had got crazy good start and was making a lot more science and culture than anyone else. So basically the game looked like either religious victory for Scythia or culture/science victory for Kongo.

I had got my commercial hubs + hansas combos up so my production was decent at this point. I desperate needed more cities for so I planned to attack Norway when I get muskets and bombards made. Researched gunpowder and no niter found anywhere close and no-one willing to trade it. So I tried with now very old catapults, knights and crossbows but city defenses where too much for those at the point. Extra bonus was of course that every one thinks I'm a warmonger even though I used casus belli.

Next option was to try to get to artillery + infantry as fast as possible so I did. Tried third war against Norway and it started nicely. The only problem was that Kongo decided at this point that he had enough and declared colonial war against me. Yes, he was that much ahead in tech. I checked and he had just finished satellites when I just finished flying. I got one city from Norway when it was time to turn my eye on the other border and try to figure out how to kill mech infs with infs, field guns and artillery. Luckily AI again only run around my cities, did not attack any of them and let me slowly kill the units. After that it was peace time. At this point Scythia had 7/8 nations under her religion including me. Kongo had started space race projects and had decent culture going and I was pretty much sure I will lose the game because I was still 1.5 eras from space race projects.

How the game ended? In the end Scythia still had 7/8 nations under her religion. Kongo was at 220/370 tourism. Kongo had stopped space race after earth satellite and I just finished my Mars projects and won science victory. Getting my first deity victory this way did not feel that good. I should have lost that game.

 

[fatgordy]

Getting my first deity victory this way did not feel that good. I should have lost that game.

It seems like the AI starts heading in the right direction and then easily loses focus on achieving their goals. Reminds me of my kids when I tell them to clean their room.

 

[georgjorge]

I *think* that if I play on Immortal and get joint declared on by Gandhi and Gilgamesh, both my neighbours, early on while I have nothing more than two Archers and Gilgamesh alone has at least 12 of his Chariots...

...then I shouldn't be able to defend myself just by taking shots at Chariots, who then turn back damaged, replaced by a new Chariot, which gets shot at again, and so on. After 15 turns of this (and three Chariots actually killed), I make peace and get some gold out of it.

I see the above argument about stacks of doom allowing only limited war tactics. But I know that if I had been attacked by a 12 Chariot STACK in IV instead of this, my cities would have fallen - just like they should.

 

[Shorlin]

I see the above argument about stacks of doom allowing only limited war tactics. But I know that if I had been attacked by a 12 Chariot STACK in IV instead of this, my cities would have fallen - just like they should.

In Civ5 the AI had no problem surprising you early game and taking a poorly defended city from you. The problem in 6 is not so much 1upt as the AI's seeming inability to effectively attack a city.

 

[TurboJ]

In Civ5 the AI had no problem surprising you early game and taking a poorly defended city from you. The problem in 6 is not so much 1upt as the AI's seeming inability to effectively attack a city.

Or attack a unit. Or build any naval units. Or build any air units. Or... In Civ V the shortcomings of the AI "intelligence" were offset by unit spam. Now the AI doesn't build units. Sometimes they build land units yes, but even those are 95% incapable of doing anything useful. When playing Civ V deity, the enemy had defensive artillery. They had 10 fighters per city to stop your bombers. They had battleships and destroyers who at least slowed down your navy some.

Now AI just doesn't do war at all. Neither do they production. Because of this, AI can only win by tourism or religion, and even that they can only do if you don't try to stop them.

I don't see how all the pro reviews of Civ VI have been silent about the lack of AI. And here's the thing - it wouldn't matter to me if the game was bad or mediocre. But the game itself is great. It feels like it could be the best 4X ever. Why can't it have even Civ II's worth of AI? If it's so damn difficult to program such a thing, then make some scripted beelines and forced quotas of different military units. And making the units sometimes ATTACK for crying out loud... It's frustrating me because how wonderfully "next level" Civ VI is on nearly every other aspect. Now for the first time I abandoned a game of Civ VI and the only reason I can think of to play any more is 100% role playing. I seriously hope they will give the AI some attention soon.