Is the AI really that bad compared to Civ4?

Gort said:

I don't really agree with this. The AI can only win a diplomatic victory by accident (EG: Greece buys up city states throughout the game, and will eventually win a diplomatic victory due to that, but doesn't buy up city states specifically to win a diplomatic victory vote), doesn't know that it needs to capture capitals to win domination victory, doesn't know how to win a Tourism victory, and doesn't particularly beeline science victories and just sort of stumbles into those.

The main way an AI wins in Civ 5 is by making the player quit. Otherwise, the only thing they do is stumble into diplomatic or scientific victories.

Click to expand...

I think the AI is generally programmed not to aim for domination, since realistically an AI player can never win a domination victory (since in practice that usually entails killing the player, and the game ends as soon as that happens). Though I have seen it a couple of times, the AIs that focus on domination seem programmed to destroy civs in their entirety in turn.

As for the others, I'm afraid that in my experience you're simply wrong. AIs will certainly take specific tech paths that facilitate science victory if that's their goal, and a late-game pattern of seizing available city states is common, not just for civs that have a CS focus (a focus which, in any case, doesn't usually much affect an AI civ's strategy in relation to city-states). Civs aiming at a tourism victory actively push high tourism and Wonder-spam from early in the game. If anything, one of the AI's weaknesses is that it's railroaded too heavily towards a specific win condition chosen close to the game's start, and has difficulty adapting if this ceases to be feasible.

The ai in civilization 4 is also challenging sometimes.. the higher difficulties make the ai spoiled in technology so bad that even large stacks of less advanced units can often fail in conquering or achieving a domination victory in bts... civilization 5 ai is not as difficult as bts ai though since I was able to be victorious in immortal difficulty twice already and the rest of victories in emperor. In bts I've only achieved a monarchy victory and it was domination once.. I couldn't believe it but then I tried monarchy again and I wasn't so victorious anymore.. I feel like if I had a prime in bts that lasted for a temporal time... The first civilization ai is really spoiled but then the ai could spoil if you take its cities. Sort of like having the treasures of Nineveh unique attribute that steals technology upon city capture but with any civilization.

PhilBowles said:

As for the others, I'm afraid that in my experience you're simply wrong. AIs will certainly take specific tech paths that facilitate science victory if that's their goal, and a late-game pattern of seizing available city states is common, not just for civs that have a CS focus (a focus which, in any case, doesn't usually much affect an AI civ's strategy in relation to city-states). Civs aiming at a tourism victory actively push high tourism and Wonder-spam from early in the game. If anything, one of the AI's weaknesses is that it's railroaded too heavily towards a specific win condition chosen close to the game's start, and has difficulty adapting if this ceases to be feasible.

Click to expand...

So, you haven't regularly seen AI civs lose a diplomatic victory with thousands of gold in their back pocket?

I think you've too high an opinion of the Civ 5 AI regarding the other victories as well. Civs that like to build wonders will build lots of wonders, regardless of what victory they're pursuing, and they won't synergise their social policies to go with it. It's not them going after a victory, it's just them having high flavours for wonders.

I can't recall ever seeing the AI win a tourism victory, though I'm sure it eventually would given infinite time.

I've seen 4 'let's plays' by Marbozir, 2 conquest/domination victories (Zulu + Morroco), 1 science victory (Rome)
and 1 cultural/tourism victory (France), it simply looks bad.
Even with a 10 tech lead around the industrial era + number one in science or having a massive culture output,
the AI isn't able to trigger a win (let's play civ5 as France).
I tried to watch him playing Carthage on Deity, but as soon as he talked about the map type, Archipelago,
you know which way the game was going.
All capitals coastal means a rush towards battleships and game over.

PhilBowles said:

I think the AI is generally programmed not to aim for domination

Click to expand...

I have never heard of game where the AI won by DV, but they do plenty of steam rolling. They also start building SS parts, so SV happens first.

since in practice that usually entails killing the player, and the game ends as soon as that happens

Click to expand...

No, the game will keep going. I wish there were a specific mechanism for observing though.

If anything, one of the AI's weaknesses is that it's railroaded too heavily towards a specific win condition chosen close to the game's start, and has difficulty adapting if this ceases to be feasible.

Click to expand...

I feel like the AIs are actually not focused enough. If the runaway is close to CV, why does it also keep building SS parts?

Gort said:

So, you haven't regularly seen AI civs lose a diplomatic victory with thousands of gold in their back pocket?

Click to expand...

My understanding is that the AI is programmatically limited in ways they can spend their money. They won’t cash rush buildings either. I think this is overall good game design because otherwise the advantages the AI gets would make them insurmountable because of the way the gold gets compounded.

It's not them going after a victory, it's just them having high flavours for wonders.

Click to expand...

I agree that this is mostly true. It actually seems kind of remarkable to me that the flavors synergize enough that the AIs don’t flounder more.

I can't recall ever seeing the AI win a tourism victory, though I'm sure it eventually would given infinite time.

Click to expand...

Now I have to ask, what level you are playing at? At Immortal+ it very common in my games to one or two AI civs running away with tourism towards a CV. This is good competition as it provides time pressure on me to get my SV done quickly enough. Frankly, I still don’t understand how the AIs are generating so much tourism, because the wonders are spread out, and this is without France or Brazil in my games.

Military ai is utterly unable to win wars. If there are hills of forest or something that does not allow more than one move per turn you cant lose against ai no matter what odds. Diplomacy ai is equal bad as well, like Holland ai most likely dont understand how to use it UP. I see civ5 as a failed game because ai simply fails too much and on higher difficulties ai with all bonuses just make no sense.

Gort said:

So, you haven't regularly seen AI civs lose a diplomatic victory with thousands of gold in their back pocket?

Click to expand...

There's a difference between an AI not trying to win a particular type of victory, and an AI not using exploits a human would use for the same purpose. I believe people who've looked into the code have found that AIs actually have a specific limit on the amount of gold they're allowed to spend to bribe CSes every turn (or on the number of CS allies they can bribe a turn).

Gold-harvesting CSes is a human exploit which is an effective way to quickly switch to and win a diplo victory you hadn't previously planned, but it's not an efficient way to gain CS favour in the longer term. If an AI has 'win diplo' as its favoured condition from an early stage in the game, and starts accumulating city-states to that end, gold is a very poor way to win diplo victory - espionage (which AIs very often use to retain CSes), Patronage (which diplo-focused AIs will always take) and quests (which they seem to pursue less actively) are much better.

And, as I mentioned, the Civ V AI is highly inflexible - if it has a lot of gold but 'win diplo' is not its chosen victory condition, it's not able to adjust to change its victory condition.

I think you've too high an opinion of the Civ 5 AI regarding the other victories as well. Civs that like to build wonders will build lots of wonders, regardless of what victory they're pursuing, and they won't synergise their social policies to go with it. It's not them going after a victory, it's just them having high flavours for wonders.

Click to expand...

Yes, civs with high Wonder flavour will build a lot of Wonders. That's an entirely separate issue. Often these civs are primed to favour going for a cultural victory, but as you say they won't always pursue it. But any civ will spam Wonders if it's chosen a cultural victory focus, including civs that otherwise don't particularly prioritise Wonders, and they will also often build the appropriate Wonders.

It's true that AI social policy selection is generally poor, but you are I think confusing the fact that AI civs don't play the optimised paths to victory humans do with the claim that they aren't aiming for a specific victory at all.

I can't recall ever seeing the AI win a tourism victory, though I'm sure it eventually would given infinite time.

Click to expand...

I've seen it at least once on Immortal, and on quite a number of other occasions they've come close. Certainly immediately after BNW was released most civs - even science-focused ones - seem to have been programmed to favour cultural victories (which was very bad for gameplay since the nature of cultural victories means that only one or two civs will ever have meaningful amounts of tourism, so most AI civs in any given game would end up wasting a lot of resources).

beetle said:

I have never heard of game where the AI won by DV, but they do plenty of steam rolling.

Click to expand...

I've certainly seen cases where the AI seems more systematic than simply 'steamrolling', focusing heavily on taking down one civ, and then the next, and then the next, much as I've observed in Civ IV when Genghis, say, will subjugate or destroy each neighbouring civ in turn. This is different from the usual 'spam army, check diplo to see who I hate most regardless of where on the map they are, CHAARGE!' behaviour that's more typical of Civ V AI.

I feel like the AIs are actually not focused enough. If the runaway is close to CV, why does it also keep building SS parts?

Click to expand...

I agree, and I think this is programmed as it is because the AIs lack flexibility. That way, if they do fail at their primary VC they can 'stumble' into another in much the way the earlier poster suggested. Whereas a player would abandon a VC they can't win and change focus, the AI can't do this so has to try doing everything in parallel.

That said, I have seen some examples of unexpected AI adaptability to circumstance. I recall one game where I overtook Germany, then going for SV, in science; my spies in German cities saw Bismarck switch all his production to science, and he eventually took the lead again.

AI is somewhat better in G & K than in Vanilla, no?

Someone mentioned some recent patch that allows AI to use excess funds to build large armies in G & K. Emperor and Immortal conquest wins are considerably more difficult in G & K than in Vanilla. When I first started G & K, I lost 3 conquest games to AI because they were able to build the spaceship a few turns before I captured their capitol.

Tiber4321 said:

AI is somewhat better in G & K than in Vanilla, no?

Someone mentioned some recent patch that allows AI to use excess funds to build large armies in G & K. Emperor and Immortal conquest wins are considerably more difficult in G & K than in Vanilla. When I first started G & K, I lost 3 conquest games to AI because they were able to build the spaceship a few turns before I captured their capitol.

Click to expand...

Yes, it's quite significantly better in G&K than in vanilla.

Sadly, it seemed to take a step back in BNW - a combination of additional modifiers and variables to keep track of in diplomacy, overweighting in the code of the diplomatic impacts of ideology and trade partnerships, problems with the way it interacts with the World Congress, and (though I think this was addressed to some degree with patches) a tendency to prioritise the expansion's new culture focus even when that's not the best choice for that civ or its strategy.

PhilBowles said:

I've certainly seen cases where the AI seems more systematic than simply 'steamrolling', focusing heavily on taking down one civ, and then the next, and then the next, much as I've observed in Civ IV when Genghis, say, will subjugate or destroy each neighbouring civ in turn.

Click to expand...

This is exactly the behavior I have observed, and I what I would characterize as “steamrolling”. A human player would usually stop or slow down after taking the cap. The AI is much more blood thirsty!

PhilBowles said:

Sadly, it seemed to take a step back in BNW

Click to expand...

I agree with this observation too. I stayed on G&K longer than most, and found the BNW pre-Ideology game to be much more peaceful!

beetle said:

This is exactly the behavior I have observed, and I what I would characterize as “steamrolling”. A human player would usually stop or slow down after taking the cap. The AI is much more blood thirsty!

Click to expand...

I wonder whether the domination AI is simply unchanged from Civ IV, where this was the rule to follow. In any case, it works better for the AI - an AI that tries for capitals only is likely to be prone to ambush on the way if the player has intervening cities.

I definitely think there's a difference in this behaviour from that of an AI that might be aggressive but doesn't actively have domination as its chosen victory. So even though an AI can never get an (observed) domination victory, you can infer that it's shooting for one.

I agree with this observation too. I stayed on G&K longer than most, and found the BNW pre-Ideology game to be much more peaceful!

Click to expand...

That may be to the AI's advantage since it can't fight against higher-level human opponents effectively. But it stumbles diplomatically once ideology hits and old game-long relationships become subject to the arbitrary whims of which ideology has been chosen, AIs select ideologies suboptimally, and their efforts at World Congress resolutions rarely advance their interests except for science vs. arts funding. It may be less a case of the AI becoming weaker and simply that it's almost entirely incapable of using the new game mechanics BNW gives the player; the result is essentially the same.

Gort said:

So, you haven't regularly seen AI civs lose a diplomatic victory with thousands of gold in their back pocket?

I think you've too high an opinion of the Civ 5 AI regarding the other victories as well. Civs that like to build wonders will build lots of wonders, regardless of what victory they're pursuing, and they won't synergise their social policies to go with it. It's not them going after a victory, it's just them having high flavours for wonders.

I can't recall ever seeing the AI win a tourism victory, though I'm sure it eventually would given infinite time.

Click to expand...

Well to be honest there's a balance problem there for the AI. Flavors in the game are programmed based on the assumption that things are somewhat well balanced. Which is sadly not the case.
Piety is a poor tree and rationalism a must have. Great merchants are while great scientists are a must have. But the AI doesn't know that and the flavors were not made with optimal play in mind. And that's where half of the economic problems of AI comes from in the game, the game isn't balanced and AI were never "reflavored" in accordance to it. AI will do some crap piety commerce combo and fill merchant slots.

If you play against Sejong he goes Rationalism most of the time and his UA works very well with an AI that put specialists everywhere, making Korea a usually good AI civ (if it doen't die).

And contrary to civ4 where you could change civics, the AI is stuck with many of its decisions. Also Civ4 had the best idea when it comes to GP, each one allows you to bulb different tech. I have no idea why they didn't keep this idea. In civ5 the AI will spawn some crappy merchant or bulb scientists right away.

For military, MUPT is vastly easier to make it work with an AI than 1UPT so it is not a surprise there that Civ4 high difficulties are threatening.

I don't really think Civ4 AI is really better, I just think the game is more AI friendly, having mechanics where the AI cannot go wrong, and as a result is more challenging at max levels. But since challenge is not all there is, I still find Civ5 more fun to play due to being meatier after 2 expansions despite the game no longer being a challenge just to win games.

Gort said:

I can't recall ever seeing the AI win a tourism victory, though I'm sure it eventually would given infinite time.

Click to expand...

I literally just lost a Diety game to Austria via culture on turn 316. Was nuts.

I was the last holdout as 4 city Autocracy Shaka -- everyone else on the map was order. And Austria didn't like me much so they didn't have trade routes/diplomat -- otherwise I would have lost like 30+ turns soon.

Yes, the AI would have won around turn 250-275 with Culture if they had used trade routes/diplomats on me.

Acken said:

And contrary to civ4 where you could change civics, the AI is stuck with many of its decisions. Also Civ4 had the best idea when it comes to GP, each one allows you to bulb different tech. I have no idea why they didn't keep this idea. In civ5 the AI will spawn some crappy merchant or bulb scientists right away.

For military, MUPT is vastly easier to make it work with an AI than 1UPT so it is not a surprise there that Civ4 high difficulties are threatening.

I don't really think Civ4 AI is really better, I just think the game is more AI friendly, having mechanics where the AI cannot go wrong, and as a result is more challenging at max levels. But since challenge is not all there is, I still find Civ5 more fun to play due to being meatier after 2 expansions despite the game no longer being a challenge just to win games.

Click to expand...

I totally agree; Civ4 is a more AI friendly game. The point about changing civics is especially true, and the mechanics of moving and fighting with a 1upt army is much more complex than the AI can handle. Overall I am not sure that I find Civ5 to be more fun. I'm playing at a lower difficulty level, and still find global happiness too constrained (still on G&K). Using the sliders and civic choices in Civ4 feels more flexible.


Sent from my iPhone using Tapatalk

Gort said:

So, you haven't regularly seen AI civs lose a diplomatic victory with thousands of gold in their back pocket?

I think you've too high an opinion of the Civ 5 AI regarding the other victories as well.

Click to expand...

I believe the AI is programmed not to try too hard to win. Often a runaway with tens of thousands of gold to buy up CS (or even space ship parts) and enough production to overwhelm you with numbers will lose. This is not so much bad AI because it is deliberate.

Gort said:

Civs that like to build wonders will build lots of wonders, regardless of what victory they're pursuing, and they won't synergise their social policies to go with it..

Click to expand...

True. They will even build wonders while you are taking their cities. Ramses built two wonders while losing a war to me. He completed the last one the turn before he lost his OC.

qemist said:

I believe the AI is programmed not to try too hard to win.

Click to expand...

This is certainly the case. They actually have a stat called "victory competitiveness" which sets how much they try to win. Some leaders just don't care about victory.

Which is why I always end up scratching my head when someone comes out with a line like "The problem with Civ 5 AI is that it's programmed to win at all costs!"

vorlon_mi said:

I totally agree; Civ4 is a more AI friendly game. The point about changing civics is especially true, and the mechanics of moving and fighting with a 1upt army is much more complex than the AI can handle. Overall I am not sure that I find Civ5 to be more fun. I'm playing at a lower difficulty level, and still find global happiness too constrained (still on G&K). Using the sliders and civic choices in Civ4 feels more flexible.


Sent from my iPhone using Tapatalk

Click to expand...

Well fun is subjective obviously, for example I consider happiness management one of the core thing that make Civ5 interesting Well I still have balance issue with the whole thing but as a concept I like it. I'd just wish there was a mechanic that slows down city pop locally on top of it, health for example.
For balance issues I know there are some mods but last time I looked into it, it was almost an overhaul of the game rather than a minimalist approach. Also there are some optional mods for it so it's very confusing to understand (and install) what kind of balance the mod developers are trying to achieve if it's with or without these optional mods... sounds like a nightmare to try to balance a game with a modular approach to it.

yep... (sigh) afraid so

Yeah, I was into the community mod thing until the projected started creeping with a very complex and hard to balance overhaul of the happiness system. The happiness system is one of the things I like better in civ V than in civ IV as there is less micromanagement, interesting trade-offs and an elegant solution for golden ages.

Regarding the OP. The game is harder on the AI but it is also the case that the AI code is sloppy and some of the mods to it has made significant progress with little effort. If you come form civ IV, civ V might be encapsulated in the phrase "brilliant ideas - sloppy execution".