The AI Is Still Bad?

[timmychompchomp]

I released an AI mod. I think it's quite good.

Thank you for sharing this and for the effort. Did you send a report to Firaxis about the bugs you discovered? If not, please do so.

 

[Ornen]

My experience on Sovereign as well, @tedhebert (in fact, Amina does almost too well there for turn 107). There must be something that throws the AI off in some cases, could be contraintuitive - maybe higher difficulties, a certain map type, whatever.

it's very strange how inconsistent the AI is... looks like they're doing really well in that game, much better in my latest. (surprisingly, the best I've seen the AI do was my first game, at level 3 / viceroy)

I've also wondered about difficulty & map types... my theory was that they do better at continents plus, that example looks like fractal though so idk. they don't seem to handle archipelago well, from what I've seen

I'm also starting to wonder if the other civs are suffering to the IPs at higher levels.

 

[tedhebert]

it's very strange how inconsistent the AI is... looks like they're doing really well in that game, much better in my latest. (surprisingly, the best I've seen the AI do was my first game, at level 3 / viceroy)

I've also wondered about difficulty & map types... my theory was that they do better at continents plus, that example looks like fractal though so idk. they don't seem to handle archipelago well, from what I've seen

I'm also starting to wonder if the other civs are suffering to the IPs at higher levels.

Yes I confirm it's fractal... really fun map actually

 

[Melkus]

End state of my first full game, also Fractal (with a similar stamp-shaped sea even) + Sovereign. There's less land = might be less room for IPs? Who knows.

I wouldn't say it was brutally hard - I basically won twice while experimenting with some stuff and learning the mechanics, Economic Victory is obviously a click away and the World's Fair will be finished in three turns - found exactly 15 artifacts btw - some of them while the AI already parked on them. So Culture was an exciting race, the world tour of my banker less so (basically a turn timer with added clicks, though it has some charm to say goodbye to everyone - my gold pile proves that I basically rushed through the last 10 turns).

But: Himiko had started work on the victory spaceship already (while being Communist and starting a massive war against democratic Machiavelli - I avoided ideologies and the Alliance she offered to stay out of it). So I won by about 10 turns, and she would definitely have won on her own terms (though I suspect she fumbled the projects a little bit, her science lead was huge). Nobody got factory resources going, that might be a weak area for the AI - or they simply weren't specialized for it.
We both finished with about 20 legacy points, I think I pulled ahead by one with the Supersonic flight at the very end.

Of course, many players would have crushed Himikos coastal empire with military and superior tactics early on, but I wouldn't say the AI is hopeless if it's capable of this at launch.

 

[d4everman]

I don't know what 1 and 0's the AI has to crunch to figure out where to settle, but it's borked. It makes no sense. (and this seems to be something that happens after the patch on release day.) I posted earlier how Rizal tried to settle a spot that was just silly. After that I found a natural wonder right next to his "area". I kept thinking "He'll grab that before I can build a settler and nab it.". Nope...he built a settler and sent it WAY UP NORTH next to Augustus.

Augustus and Catherine are both at the northernmost parts of the continent. (East and west, respectively). Rizal ignored resource rich areas between his capitol and Augustus to settle by him. WTH? It gets better...I had taken my area of the continent and was almost at my city limit (7/8). The crisis starts. It's barbarians, so I had to go squash them before they got too powerful. What do I see? Augustus sends a settler heading right for my northern border. I don't care if he built a city there...but he didn't. He ran right down next to my capitol and between three cities to one of the three green hexes there. It made no sense! There were literally 4 maybe 5 better areas he could have near his own empire. O was so pissed I had to go back to an autosave where I saw his settler, buy my own settler and plop it on that spot before he got there. Then he sends his settler up north and the dude just sat there not moving until the end of the age. Meanwhile Catherine sent a settler to the spot Augustus SHOULD have built in an established a city.

Man, they need to fix this. It's so annoying and stupid I can't believe anyone thought this was a good idea for AI settling behavior.

 

[Taefin]

In my current fractal deity game, AI was peaceful in antiquity, I built a bunch of wonders, started forming DL settlements, and now on turn 33 four AIs are at war with me, threatening my two HL towns. I’ll report back. It’s possible this happened because half my army was island hopping to clear room for DL cities and I only had 5 units at home.

 

[Ornen]

I released an AI mod. I think it's quite good.

first impressions: this seems to be the difference maker. in the two games I've tried so far, AI is actually putting in the effort. there is enough resistance here it feels like I'm actually playing something, rather than just building and building. anyone disappointed in the AI needs to give this a shot.

 

[Pfeffersack]

1. Overall the goal of being competitive with minimal bonuses is wrong. AI is just one of the game mechanics to provide challenge and how interesting the challenge is doesn't depend directly on whether it's done through more bonuses or more complex logic. So, all those mentions of "good" AI should be made in quotes.

Strongly disagreeing with this...or better said: For me it makes a difference from what source AI competitiveness comes. It is same the reason why a lot of work wents into programming chess computers...applying your logic, you could also achieve a competitive opponent by e.g. giving the chess AI a fresh extra rook on its starting tile as soon as the original one is moved. You are of course free to enjoy that kinf of "competitive AI", but it isn't fun in my book.

 

[Gehennas]

After starting a new game, I think I started noticing some patterns with AI issues.
AI really loves wonders. Probably more than everything else. I am couldn't build more than 3 wonders so far, because AI just builds them right away.

This probably it also depends on the "chosen" age focus of the AI.

In my current game, still in antiquity, only Franklin has settled up to a limit. Guess what? He's the only one not busy with spamming wonders.

 

[notque]

Yes, flavors play a key role. But there are tons of real AI issues going on, and that's why I have made an ai mod.

 

[Melkus]

Prioritizing wonders is competitive - it's literally one of the winning conditions for the age. Its a risk because maybe the AI is getting attacked - but maybe it's not. Admittedly you have to keep them, but they have to make the game work for peaceful play (huge difference to a pure strategy game like chess).
That's an interesting balancing issue, especially with so few opponents in Antiquity.

(Optimizing the settling logic is almost always a good thing of course, as long as there's some variance)

 

[Noble Zarkon]

It takes a long time to settle and expand, and tends to not have a coherent empire

I definitely agree with the second point, they need to prioritise settling nearby to trigger their road.

The first, in my observations about half the AIs seem to manage to their cap - the others may get caught up with IP issues etc (or chasing Wonders). Overall the AIs are building much bigger empires in my experience.

 

[Elhoim]

I definitely agree with the second point, they need to prioritise settling nearby to trigger their road.

Yeah, from what I can see some Civs have some settling preferences that kinda borks the settling (with Egypt doing some very weird extreme coastal settling).

 

[Melkus]

Another fractal map game, another decent (though not great) AI performance (on Immortal):

Met Isabella of Egypt very early and had to take karmic revenge and take her second city, burning arrows are fun. Her AI really had a hard time attacking over water - not news, but at least she built like six Medjays - before I burned most of them. She had some really rotten luck with two IPs to the South and Tiwanaku, they constantly attacked her somehow.
Some settler silliness visible, unfortunately. Waiting for her to build me a wonder now (and for a higher settlement limit).

I think the tactical AI is no match in this situation even for a mediocre player like me without extra settlers / units at the start. I got lucky with the map and IP situation though.

But the two (sort of) Germans are doing pretty well. Frederick has been communicating nonstop since we met, sending endeavor after endeavor.
Some people here really seem to play a different game (or the patch has improved things? - nothing in the patch notes)

 

[Mr Jon of Cheam]

Yes, flavors play a key role. But there are tons of real AI issues going on, and that's why I have made an ai mod.

Petition to have FXS incorporate all changes in the full game so that I can try it (PS5).

 

[Taefin]

Based on my first game on deity (here turn 1 of modern), the “cheats” do create an engaging game, with two AIs threatening to run away with it and making me think how I want to counter them.

When I was at war with 4 AIs from my continent (some were allied) , both distant AI (Amina and Napoleon) sent a threatening number of units across a fractal sea, both to capture a town closer to them, and to threaten my main cities far from them. The two neighboring AI (Confucius and Friedrich) made strong attacks, each taking a town and threatening my cities. If it wasn’t for ridiculously good natural barriers, and getting cogs back and unit upgrades just in time, I could have lost both cities.

The +8 CS on top of them having strong UU (chevalers) made it challenging, and I lost a commander to archers during a rout, so I didn’t have commander CS buffs fully up (the +5 in radius would really help!). I had +4/5 war support against my main enemies, and Mamlukd getting +12, and it was still all about applying chip damage with archers to give my melee a chance to get a minor victory.

Tactically they weren’t great, but did pull back or retreat when the tables turned, giving a feeling of responsiveness, while switching to defense before they lost their army. The new surrender logic is better, with Friedrich giving a town back that he captured from me, but only after I had it surrounded (Amina refused to end the war on even terms until suddenly she offered a town, I gave it back since it shouldn’t have been offered). Still saved me to end that war with Friedrich a turn early to focus a group of knights who had crossed the continent to get me, so I like that giving towns in surrender is part of the game, if better tuned.

I’m excited to try the AI mod, this game just happened to not have as many bad forward settling locations, which it addressed, but Friedrich really struggled to keep his island towns safe from IP. The AI also occupied but did not disband an IP a few times, but disbanded them other times. Napoleon was coastal and did expanded across the ocean. My last two settlements were placed in the last turns of exploration, so I would say settlements were even, and I was expanding as fast as I could while fighting that brutal defensive war.

 

[Lobox]

From what i've seen the AI seem semi-competent when using their armies, which combined with the combat bonus they receive on high level should make them a decent threat to most players on higher difficulties.
Economically, they still receive plenty of yield multipliers on higher difficulties so that should also be OK.
They appear to have a tendency to settle really odd spots but that's probably something FXS can fix relatively easily with some tweaking of a few weights, so i wouldn't worry too much about that.
The AI probably won't be a threat to the best players but i don't think it should necessarily be. The AI isn't an actual opponent, it's not there to win but to make the game interesting. If it forces you to be smart, then it has achieved it's goal.If you can win while rolling your face on the keyboard, then it failed.
Obviously we'll have to see, i had some real concerns at some point, but after seeing more videos and listening to other players, i think it's not that bad. Obviously it's not great but it will improve over time, it always did in previous titles. Hopefully it will be good enough to fulfill it's role, but players looking for an actual opponent should play multiplayer as with previous titles.

AI in CIV5 and CIV6 is still incompetent. Firaxis obviously have problems to recruit competent AI programmers or they just doesn't care.

 

[Melkus]

Waiting for her to build me a wonder now (and for a higher settlement limit).

And here I am, slogging through some fairly resilient defenses. Isabella's last stand ultimately was doomed of course (no idea why her best horse is taking a bath in the middle of the fight btw), but this sort of campaign takes much longer than it would in Civ6.

In my blind hatred of Isabella (roleplayed this a tiny bit) I was running out of time to build up infrastructure. The age ended quite early (around turn 120), too, thx to science pressure from Charlie, so I didn't get much done in terms of actually scoring (1 legacy point each).

Also, Isabella's allies could have given me a lot of trouble in the north, but preferred not to. So disappointing if you were rooting for the AI.
My biggest AI peeve though: there seems to be basically no commander logic for the AI, they run around all over the place - with the IP crisis the game turned into a commander obstacle course at times.

Ending on a positive note: observe the shattered settlement limits of my other opponents. (Frederick was clearly overstretched and went down to 6 though in the following "Rise of the IPs" crisis). Looking forward to drown everyone in culture with Hawaii now.

 

[Piranga]

Why is the AI settling crap settlement that are completely out of their way to defend when there are great city locations directly next to their capital? Every game, most AI's found these absurd settlements.

Napoleon and Tubman both with cities north of me in pure ice with minimal resources. (I killed multiple of Tubmans other settlers up there as well).
Then check out the lands right near their capitals. Tubman has so many spots available, all with resources. Napolean is a bit more crowded but there's a great 3rd city in their he's skipping. (Napolean does get some of the GBR but its still not an effective empire building city.

The game gets tiring when the AI's empires a sporadic cities in terrible spots.

 

[Wabnosa]

Might I remind everyone that we are in the beginning of a potential age of AI. Someone should train an open AI bot to play civ. Then make them roleplay as each leader. It's like were griping about the lack of Iron at the beginning of the Iron age. Firaxis has the money to do it.