Civ VI AI not THAT bad

[ShakaHulu]

I see so many posts just completely savaging the AI in this game. Personally, I don’t find it THAT bad, and here is why.

The fact is most wars that we players engage in against the AI are lopsided affairs, but that’s not because the AI is tactically incompetent. It’s because human players realize when we have the advantage and use that time to attack. I’d like to see how well you play if you were attacked by a numerically superior force all with higher combat strengths. Your tactics might look similar to that of the AI waging a defensive war. When left with all bad options every turn, sure the AI looks like it’s play weak. But it’s not because it’s tactically inept, it’s because human players are aware of when they have a significant number/technological advantage and use that time to strike.

Where the AI is truly weak is not in its tactics, but in realizing when they have an advantage and using that window of time to strike. They also tend not to spam build units during war time, which is obviously a mistake. Strangely enough it’s not the tactics of the AI which fall flat, it’s their production priorities and failure to use technological advantages when they have them. One would think these would be rather easy flaws to remedy? When at war spam build units. When you have a superior military tech use that window to strike.

These failings are what give the AI the appearance of tactical ineptitude. Honestly their unit to unit tactics are OK, though certainly not great. But when an intelligent human player uses a window of opportunity to strike it makes the AI look stupid, when really their armies are just out classed and facing a battlefield with few good options.

The thing is I’m sure the AI has on occasion done the same thing to you. But when that happens to a human player they simply end the game, start a new one, and chalk that game up to a bad start or some other anamolous fluke. Soon forgotten! New game! And then the next game when you crush the AI and mourn how woeful they are, forgetting all about that time they actually got you. Mind you this scenario doesn’t happen often if you play on a difficulty level commensurate with your skill. But play on deity where the AI has enough production to overcome their poor production choices and see if the AI is incompetent when they have an army that is superior to your own.

TL;DR - What often appears as poor AI combat choices is really just a human players army outclassing the AI’s army. The real flaw in the AI is not their battlefield tactics, but their failure to build enough units or to attack when they have tech superiority. Oddly enough, these should be relatively easy fixes.

 

[Zike]

They really are absolutely trash across the board.

You can win a war when the enemy has like 5 times your army size. In fact I have.

 

[Hammurabbit]

I see so many posts just completely savaging the AI in this game. Personally, I don’t find it THAT bad, and here is why.


Where the AI is truly weak is not in its tactics, but in realizing when they have an advantage and using that window of time to strike. They also tend not to spam build units during war time, which is obviously a mistake. Strangely enough it’s not the tactics of the AI which fall flat, it’s their production priorities and failure to use technological advantages when they have them. One would think these would be rather easy flaws to remedy? When at war spam build units. When you have a superior military tech use that window to strike.

I agree that this is probably where the AI is the weakest, i.e. using a tech advantage / production advantage in a military campaign against you. I think the only time I really see them try is in the ancient era, where the AI very often try to attack you with a larger army before you have built up your own force. On King and higher they are using their extra settlers and production to take you on early. I also consider these the most interesting wars, and you really have to use your human brain to protect yourself from the first wave of troops. I wish the AI would do similar attempts later in the game as well, when they still have the lead in science and number of cities.

I also agree that tactically, the AI is not as smart as a skillful civ 6 playing human in managing their troops, but I think that's kind of OK, and partly why the developers have added production bonuses and fighting bonuses on higher difficulties. At least when you outsmart the AI in field tactics, beating a larger army, you feel a bit happy with yourself.

 

[plus]

agree.i have scytia as neighbour and we had endless wars..i play long games and she has so many horses....i think the ai is ok and i dont use any ai mods.

 

[sonicmyst]

I think the AI is competent if they are settled in areas with the resources they need. Look at AI civ without iron, niter or horses, and they are like sitting ducks.

 

[Lucius_]

I usually don't hesitate to attack an AI with a tech advantage and I never pay attention to their military size beforehand. It can be a slog taking their first city in such situations, but after that, it's downhill for them from there. Generally they never have a "complete" military and I've rarely seen them defend well. I agree they have a production priority issue, especially by mid and late game, they just don't make units anymore. It's like there is no programming giving them instructions to keep a standing army on hand for defense or offensive if necessary. Outside the early game, I've never had an AI show up with a mixed military capable of doing anything. The AI also needs to learn to park their units and wait for a full siege force, instead of trickling in to attack a city 1-2 units at a time.

 

[halfhalfharp]

I normally build little to none military units except for those used to fence off barbarians in the early game.
And I tend to use diplomatic skills to befriend neighbors in order to save resources something else. Therefore, when there is an unexpected DOW, (there is some bugs in alliance system that somehow allows allies to joint war lol, or some far away neighbors joint war against me), I am just unprepared.

And even using this vast gap of number for invasion, the AIs cant harm me. I just buy+all in building units, and I suppose my defensive tactics are not that great anyway. However, in almost all cases, not only I defeated them all and stopped at defending, but became the invasive force myself.

I was using crossbowmen at that point against knights, as I mostly invested the upper branch of the tech tree. If they sent in endless waves of units, my ranged units would have been overwhelmed for certainly. but they just miraculously stopped doing so at some point.

They really just cant stand a match in wars, even in Emperor levels. Immortal levels and onwards may have a difference but nothing cant be handled with an all-in military strategy.

They surely know how to take advantage of (military)weakness, although maybe not too well. Their strategies are indeed very underwhelming. That AIs do not know which war type is beneficial is an obvious example.
Cyrus likes to declare formal war. And I have never seen Tamar protecting her nearby fellow city states as well. And Australia, John Curtin always thinks that they are on a cross road. And Gilgamesh always stands alone because everyone hates warmongers.

 

[Hammurabbit]

I did have a situation in a game recently, where Mansa Musa, a modded leader that I accidentally had in my game (disabled mods are automatically enabled when they are updated it seems), did a great job managing his military.
First, he marched a big army across the map, avoiding to attack me who had decent defense and quickly sieged and took much weaker Victoria and London.
Turns out I didnt have access to horses or iron (and I was mostly living on the tundra - a really crappy start location). London happened to have the only iron in sight, Mansa Musa upgraded his warriors to swordsmen and swiftly attacked me. With no decent units to defend against his swordsmen I had to give up the game.
I wonder if it was pure luck, or if the leader mod changed build priorities enough to make Mansa a better warrior AI.

 

[PhilBowles]

Where the AI is truly weak is not in its tactics, but in realizing when they have an advantage and using that window of time to strike.

Which in the context of Civ VI is another way of saying "realising they have a melee unit next to a 0 health city and using that unit to attack instead of sitting around dumbly or wandering off pointlessly".

 

[Trav'ling Canuck]

Which in the context of Civ VI is another way of saying "realising they have a melee unit next to a 0 health city and using that unit to attack instead of sitting around dumbly or wandering off pointlessly".

This may also be an AI prioritizing thing. The AI may not want that city.

Tactically, I think they do know how to get it if they want it (they may be misled sometimes if they rely on the same information the pre-combat info panel gives you, as it doesn't always seem to adjust it's prediction to reflect a higher probability of success when a city is close to zero health).

Strategically, we don't know what the AI's current priorities are. They may have launched the initial assault with the idea of taking that city. Then at some point, whether that's every turn or periodically, the AI's priorities are re-evaluated and they decide they don't want that city. Or maybe they never intended to capture the city, and it may simply attracted attacks as the highest priority enemy target, but they aren't on a "capture city" mission so don't do so.

This is back to the flexibility and taking advantage of opportunities thing. A player may recognize they have the resources available to tackle a new opportunity they weren't originally thinking of. As far as I can tell, the AI won't take an undefended city unless it's specifically on a mission to capture that city.

 

[TheMeInTeam]

The AI will occasionally not take cities with 1 HP with an adjacent melee unit. It doesn't even play civ 6, it plays something else.

"Not that bad" relative to what?

 

[Revolutionist_8]

I was surprised to see that Pedro waited with the war declaration until he surrounded my warrior with 3 warriors (and attacked it ofc with all 3), used horseman to destroy my trade routes, focused my damaged units, retreated with low hp units, and so on.

Also i've noticed that AI sometimes buys good tiles too, and not just wait for culture-expand. Furthermore seen quite decent district placements as well.

On the other side, sometimes very weird and questionable what the AI does. Not improving and repairing tiles, doesn't attack with ranged units, etc.

From my observation it needs development and tweaking (like non-sense joint war declarations), but the general direction is promising, I hope they will keep hammering it

 

[Siesta Guru]

This may also be an AI prioritizing thing. The AI may not want that city.

Then it shouldn't waste units trying to not take it, lol..

There's no point in trying to anthropomorphize it and try to assign human-like motivations, based on what's visible and the xml, these are the reasons:

- The unit is able to do an action that's higher on the static priority list. In vanilla these are on the priority list (removing some not applicable to melee units):
Leader Formation, Support Formation, Attack High Priority (units), Attack Camps, Move to Safety, Gather Goody Hut, Attack Medium Priority (units), Attack Civilians, Promotion, Make Military Formation, Attack Low Priority, Heal, Pillage District, Pillage Improvement, ATTACK DISTRICT, ...
Since attack district is so far down it usually has something higher on the list that it meets the requirements for doing. Especially infuriating is how they put "Move to Safety" higher, which means that if the unit is low on health and can be attacked, it'll run away.

- The melee unit is not part of an operation that tries to attack the city and just happens to be there
- The operation the unit is part of is not in the correct state to allow melee units to attack cities

 

[Trav'ling Canuck]

Then it shouldn't waste units trying to not take it, lol..

There's no point in trying to anthropomorphize it and try to assign human-like motivations, based on what's visible and the xml, these are the reasons:

- The unit is able to do an action that's higher on the static priority list. In vanilla these are on the priority list (removing some not applicable to melee units):
Leader Formation, Support Formation, Attack High Priority (units), Attack Camps, Move to Safety, Gather Goody Hut, Attack Medium Priority (units), Attack Civilians, Promotion, Make Military Formation, Attack Low Priority, Heal, Pillage District, Pillage Improvement, ATTACK DISTRICT, ...
Since attack district is so far down it usually has something higher on the list that it meets the requirements for doing. Especially infuriating is how they put "Move to Safety" higher, which means that if the unit is low on health and can be attacked, it'll run away.

- The melee unit is not part of an operation that tries to attack the city and just happens to be there
- The operation the unit is part of is not in the correct state to allow melee units to attack cities

Thanks for setting these out. I was aware that there are priorities and missions/operation groups that drive AI behaviour, but haven't seen this set out so succinctly before.

Out of curiousity, do you know if these lists can be customized by Civ in the current implementation? I've often thought that it would be interesting if the troops of different leaders would behave differently. So, for example, one leader could have "Move to Safety" high up the list, making them a Cautious leader (in old grognard talk), while another leader has it way down the list, making them a Bold leader.

Just having a mix of AI behaviour could make war with different civs feel differently (one is more likely to group into formations before advancing, another is more likely to rush attacks, etc.) Plus it would help identify which combination of priorities is most likely to lead to optimal AI effectiveness.

 

[Archon_Wing]

The list of things the AI does right either by luck or ability is pretty small. Much smaller than the blatant things it can't do, such as something as simple as settling fresh water as shown above. In fact, it like avoids it. Or simply not bombarding units when there is no reason not to

And the AI doesn't just fail because the human outclasses it. They are also woefully underequipped to deal with barbs because they poorly escort settler, or even city states at times. They're not being outplayed by anyone but themselves

If the AI were merely tactically incompetent, I don't think it'd get as much flak. But it makes fundamentally bad decisions every turn.

 

[Siesta Guru]

Thanks for setting these out. I was aware that there are priorities and missions/operation groups that drive AI behaviour, but haven't seen this set out so succinctly before.

Out of curiousity, do you know if these lists can be customized by Civ in the current implementation? I've often thought that it would be interesting if the troops of different leaders would behave differently. So, for example, one leader could have "Move to Safety" high up the list, making them a Cautious leader (in old grognard talk), while another leader has it way down the list, making them a Bold leader.

Just having a mix of AI behaviour could make war with different civs feel differently (one is more likely to group into formations before advancing, another is more likely to rush attacks, etc.) Plus it would help identify which combination of priorities is most likely to lead to optimal AI effectiveness.

Yes, you can shuffle the priority list around in XML and have different lists be attached depending on some factors. In AI+ have a slightly different ordering on deity than in other difficulty levels, leaders are also possible.
But, you cannot order it based on smart in game factors, only those conditions available in XML (and you can't mix & match well, so no separate version for montezuma deity)

 

[PhilBowles]

This may also be an AI prioritizing thing. The AI may not want that city.

In which case it should be coded to raze it (as far as I can tell, AI civs never raze cities, only city-states do - and that because they're coded to be unable to keep a captured city), and shouldn't invest effort in attacking it if it isn't interested in doing either. What's more the AI itself gives the lie to this reasoning - it will frequently vacillate and wander around the city for a few turns, and then capture it several turns down the line - or make an effort to, often after there's been time to get a ranged unit or wall, or to bring up reinforcements.

Strategically, we don't know what the AI's current priorities are.

No, but we can make common-sense inferences of what their priorities ought to be and compare with the other games. Civ V AIs would consistently capture cities they were able to and would rarely attack with forces that couldn't. They got stymied mainly by losing their melee units and then being unable to redirect their ranged force or to recognise that they needed to bring new melee units down. I've seen AI attacks in Civ VI where a city is attacked only by a horde of archers, without any melee units anywhere nearby.

There isn't any good reason for Civ VI's AI priorities to be any different (though as above Civ V AIs at least had the capability to raise cities they don't want), and moreover there isn't even a penalty in Civ VI for having extra cities - there are far fewer scenarios where it's bad for the AI to keep a captured city than there were in Civ V. Loyalty should factor in, of course, but evidently doesn't - when the AI does capture cities, they're as likely to be in areas it can't keep them as not, and frequently moreso since it's coded to favour attacking capitals and city states even when these are deep in another civ's territory.

 

[Eliminator_Sr]

I've put in easily 500 hours into the game mostly at high difficulty and the AI has taken 2 cities from me in total. One of them was a warrior rush on my capital in the first 20 turns. If the AI was even remotely competent at tactics they should be easily capturing my cities when I build close to no military units. Instead they just move their units around like they are fighting an invisible army while I pot shot them with ranged units 10 times in a row. Even when they manage to get my city into the red by some miracle of chance they STILL refuse to capture the city more often than not. It's so bad that I don't even bother to defend my cities anymore. Even with the AI+ mod they still make glaringly obvious tactical mistakes. People are completely justified in complaining about the AI - it really breaks the immersion of the game and is the biggest flaw by far with civ 6 IMO.

 

[earlc]

I partially agree. The AI does have real issues, and some of them can potentially be fixed without sacrificing too much computing power.

Simultaneously, the people who act like it's egregiously game-breaking and endlessly whinge about it are typically exaggerating to reach their preordained conclusions or innermost desire to hate the game no matter what. You can see this same pattern here on the forums with other supposed pitfalls of the game's design.

That said, the way the AI interacts with the player has slowly been improving, with a major leap with the release of R&F. I've seen an OK attack from the AI, but most of all I've seen some pretty robust defense. They definitely produce a lot more units than they used to, wall their cities, and will generally put up a fight if they're not a tiny civ or behind on the tech tree. Which still isn't quite good enough, but it really is improving. The team needs to find budget to work on this stuff, which is why the updates are typically centered around DLC and especially this last expansion pack. And that's really that.

 

[Trav'ling Canuck]

I've put in easily 500 hours into the game mostly at high difficulty and the AI has taken 2 cities from me in total. One of them was a warrior rush on my capital in the first 20 turns. If the AI was even remotely competent at tactics they should be easily capturing my cities when I build close to no military units. Instead they just move their units around like they are fighting an invisible army while I pot shot them with ranged units 10 times in a row. Even when they manage to get my city into the red by some miracle of chance they STILL refuse to capture the city more often than not. It's so bad that I don't even bother to defend my cities anymore. Even with the AI+ mod they still make glaringly obvious tactical mistakes. People are completely justified in complaining about the AI - it really breaks the immersion of the game and is the biggest flaw by far with civ 6 IMO.

That was my experience prior to R&F.

I lost two cities - one with walls - to the AI in my first two games of R&F. Both times it was a city state ally that took the cities while my army was fighting the major civ.

There was no hesitation. It was march up with some Warriors, a couple of Archers to fire at the city, and bang - bang - bang the Warriors hammered the city, attacking even when they themselves were injured. I thought I would save the walled city, because it held out long enough for a couple of reinforcements to arrive from the front, but nope. The AI ignored the sniping fire from my army and pounded on the city and down it went.

Then, being city states, they razed the cities, so no chance to counter-attack it back.