How is the AI ?

[Victoria]

AI cheats by pulling armies out of thin air when threatened

I thought that in a couple of early battles but it can be support from out of your vision.
Then there is another tick I learnt from them, as soon as you start assimilating an independent nation, their army becomes yours, and even better, you get some great troops, not just dross.
The AI may be cheating but from what I have seen, it’s just what it does, when you kill off the initial troops it does not cheat anymore…. Indicating to me… it is not cheating.

 

[Aristos]

I thought that in a couple of early battles but it can be support from out of your vision.

Most likely that. I have not seen any evidence of cheating like the "Steam guy show", not in the open devs nor in release. FOW is a bee in this game, and I can see people used to the Civ "AI" getting fooled into believing the forthcoming battle is an easy one when in fact there are armies within range hiding in the FOW (and the AI seems to be using that, and so should the human).

 

[Taefin]

Perhaps there ought to be a drawback to just sending the citizens out there to die, there might be one but I didn't notice it. But I don't think there was anything bad with it on an in-game level.

Probably still a good use of citizens, but I believe these draw from your cities population. Might have changed since beta (or I might have been mistaken).

I guess to the AI’s credit (not that it is thinking about this) most of what it stands to gain from a given war is getting a few mills for fame stars. Given the immense food (and probably production) yield boost AI get from difficulty, this is probably actually a good use for their units. I certainly wish they were more tactics though, and I’ll avoid suicide missions myself

Ran into an awesome main-map tactical funny. I was 3 turns away from organized warfare and attacking a tiny city. We kept trading it back and forth since there were no units so the capture doesn’t cause a battle and both sides can capture in one turn. What turned the tide was we both had two armies, I used the stronger to attack their stronger, forcing a retreat. I then used the weaker army to attack the city and slowly wore it down with two Egyptian chariot archers and scouts on a hill as cover. The retreated army couldn’t join in, and seemingly they couldn’t on subsequent turns as the battle went on. Finally I got two additional units to reinforce when I got the tech, and this enabled all my units to join. Not sure why the AI didn’t do this earlier, or if they could or not. The next turn I engaged their retreating forces and killed all of them, at which time they all joined the battle, implying they had organized warfare already.

 

[Elhoim]

Well on the Steam forums a dude is showing that the AI cheats by pulling armies out of thin air when threatened, I don't know is it so for every difficulty level but I think that's lame.

I'd take what the Steam forums say with a pinch of salt, there's always people jumping to conclusions. The player AI doesn't create armies out of thin air, never saw it all my testing. Now, IP have some weird rules, so there could be a army spawn in some situations, or heck, even de-spawn. But they are a bit like the Civ barbarians in that regard.

 

[Gedemon]

Well on the Steam forums a dude is showing that the AI cheats by pulling armies out of thin air when threatened, I don't know is it so for every difficulty level but I think that's lame.

I remember how in Civ 5 AI didn't need to pay unit upkeep even when it was -2423 gold negative, which sucked too.

I would rather have AI get more production etc bonuses than just magically appearing armies.

edit: it could of course be some bug that will be ironed out, the game is only two days fresh
edit2: not just a single guy reporting this but several.

I'd like to see the proof, I've not seen that with the map revealed from satellite during two wars.

 

[Gedemon]

ho...

I think I've seen the thread on steam, if it's the one Haig refers to, it's not the AI players that are cheating by spawning armies out of thin air, people are complaining about barbarian (ie the independent factions in HK) spawn.

yes, those spawn, how else would they appear on the map ? how could they be a threat (or a help with hired mercenaries) if their cities follow the same rules as the other players (human and AI) while being temporary settlements mid-game ?

unless it's another thread ?

 

[NombreyApellido]

i had two different teams of agressive greys perform a move in the same turn AFTER they attacked me, not once but a few times

 

[Haig]

I hope the appearing units isn't true, removing fog of war with some cheat engine could help to monitor it.

 

[Gedemon]

I hope the appearing units isn't true, removing fog of war with some cheat engine could help to monitor it.

again, I've not seen that with the map revealed from satellites during two wars (that I won)

 

[Orez]

Bad passive ai, no wars between ai. This game is a citybuilder without any challenge just like Civ 6. Civ 5 Vox Populi is a much better game.

 

[AntSou]

Bad passive ai, no wars between ai.

On which difficulty? They've been fighting quite a lot on Empire in my game.

 

[Orez]

On which difficulty? They've been fighting quite a lot on Empire in my game.

Nation, Huge map, 10 AI players.

 

[Siptah]

Nation, Huge map, 10 AI players.

AI is More aggressive on smaller maps when there is a lot of border tension early on. Also, I think, the difference between Nation and Empire is quite noticeable.

 

[Orez]

AI is More aggressive on smaller maps when there is a lot of border tension early on. Also, I think, the difference between Nation and Empire is quite noticeable.

I changed the difficulty level and size of the map and it's much better now.

 

[Virdrago]

Bad passive ai, no wars between ai. This game is a citybuilder without any challenge just like Civ 6. Civ 5 Vox Populi is a much better game.

At the start menu, when you play a new game, under Difficulty there's a Peaceful Mode option. Was that set to Yes? The AI won't declare wars then.

Edit: Never mind. I missed your last post.

 

[Lightjolly]

More competent that civ 6 that's for sure, actually knows how to fight.

 

[Taefin]

Just finished my second full game (i.e. quit halfway through contemporary), this time humankind with 10 AI on Large, no new world, and there is good and bad about the AI.

The big thing I noticed so far is that the AI undervalues range (archers early game, mortars/artillery later). It is absolutely imperative to have a non-LoS ranged units. Most of the ranged units I’ve seen since release have been EU and these are game changing. Without range, the only way the AI can defend a city is by moving their army outside its wall. Range would put pressure on the attacker to build siege or otherwise find a way to damage the units in the city fast enough to win.
Without the AI having range, most of the game becomes about about abusing choke points and cliffs. They seem to have a preferred army configuration that they build blocks of (Zhou was a warrior, spear, and two chariots) and these should be changed to include a consistent portion of archers.

This plus a difficulty between civilization and humankind would go a long way finding the right challenge where it is playable, but without the game becoming too easy to run away with.

Also, it is nice that fighters default to patrol mode, instead of “do nothing” mode as in Civ6. Haven’t seen the AI support a battle with a fighter, but it’s nice to know that my air attacks will at least be countered.

Regarding combat I have four overhauls I’d like to see that I think will play to the AI’s strengths.

1. I think the game needs the option to retreat from an active battle, perhaps by moving units out of the battlefield in a way akin to how reinforcements work moving in (either the unit could disappear and all retreated units go back to their original army, or each tile a retreated unit uses could form an army that can be used until it is full). The defender might escape with some of their units by yielding their flag, but the opponent may decline to take it (I’d at least like an option for the defender to surrender the flag, ending the battle at the end game). But the attacker has no choice but to fight to the death or last all 3 or 6 game turns. Especially as battles begin to draw in most of both sides’ armies, this seems too extreme of an outcome. Miscalculating a battle will already have serious consequences without every unit being killed. Also very easy to abuse the AI.

I’d also like a surrender option where units that didn’t retreat get captured. This could give the other side the option to accept or refuse (which should increase war support for the victim as it’s people will want revenge). The net number of prisoners could then add to the war score at the end of the war allowing for greater level of concessions.

2. I think somethings needs to be done so that city walls do not need to be penetrated in battles that are not sieges. Most battles I end up in outside a city wall, including defending against sorties, end up requiring the effective capture of the city without the benefit of doing so. Without the retreat option above, this gets me of hand really easy, and can be abused by the player. Or the city walls could just not be included in these battles (except perhaps where there are garrison tiles). Historically, a sortie would take place outside the range of the city archers where the siege engines were being build, which this would emulate well.

3. I think infantry needs to be less effective against ships when on the ground. As it is, infantry are gunning down battleships better the ships can return fire. I’d recommend adding a steep penalty on these attacks. Here again I’d rather be able to retreat mid battle if my opponent brings in ships to support a coastal siege/battle than have to pull back and wait out the battle for 6 rounds.

4. Attacking and then immediately retreating from an army should not affect that armies ability to move. It is too easy to split up AI forces by attacking and retreating from one army before attacking the other. I discovered this by accident when an entire AI city got wrapped up in a battle against one of my horsemen, allowing me to capture another nearby city relatively unopposed. Perhaps it needs to be so that if two battles are triggered by the same player and have overlapping areas that these battles will be combined, preventing this abuse. This would reasonably include units who might otherwise be within range to reinforce the more important battle.

 

[Trav'ling Canuck]

My personal view on the AI in HK is that it ticks all of my key "asks" from a 4x AI:

1. Acts as a "speed bump": check. Early reports suggest this is not true in the release build for lower difficulty levels, but once you raise the difficulty level a little, you cannot play the game as a solo experience and ignore your neighbours. You need to have a strategy to deal with them, either to dissuade them from attacking or otherwise. Pretending the rest of the world doesn't exist while you build a game-winning economic engine will not work - the world will come knocking.

2. Acts as a "pace car": check [EDIT: not so fast, see comments below]. It won't take too many games at higher difficulty levels to run into situations where you need to deal with AI empires that are racing ahead of you in the fame game. There's room for improvement here - the AI could learn how to use some game systems better, and I suspect it will as the sense I get is that this is an area of current focus for the dev team. But if you want to start up a game where you're not sure if you'll win if you don't play efficiently, I suspect the AI can already give you a run at higher difficulty levels.

3. The game plays differently depending on who my neighbours are: check. The diplomacy system is rather nuanced, so this may or may not be obvious right out of the box, but the behaviour of the AI is very much impacted by their traits. So how the game plays out will differ based on who you have on your borders (although it seems like in the release build this isn't quite so true at lower difficulty levels, as the starting difficulty levels seem to have either tamped down the aggressiveness of some AI traits or generally nerfed the impact of all AI traits, not sure which it is exactly).

An additional comment on the tactical combat competence of the AI: this has come along in leaps and bounds, in part because the dev team has been conscientious about seeking out examples when the AI acted stupidly, analyzing why it did so, and teaching it to be smarter, There are still some situations that the AI struggles to handle properly, but I expect the number of such situations to continue to dwindle. If you run into an example where the AI made a bonehead move during tactical combat, I'd encourage you to share it either here or on games2gether, as the dev team in my experience has been eager to teach the AI to fight as well as it can.

EDIT: I'm seeing reports on @Krajzen 's thread that my comments re the AI being an effective pace car in item 2 may not be true as of the release build. It certainly was true in pre-release builds I got to try out, but perhaps something broke and/or the play balance was tilted towards making things easier for the player (either purposefully or accidentally). Anyway, I'll now consider that item to be not checked.

 

[fortydayweekend]

AI is streets ahead of any Civ game on release .

Economy - district placement is a hard problem and the AI cities don't look that great to be honest. I've only played a couple of games but they don't seem to be spamming Unique districts either, which feels to me like the most important part of the game. I'm finding I prioritise building those 1-per-territory districts in almost every territory, some of them are incredibly powerful.

City Center building yields are all pretty simple to calculate so presumably it's building the best ones first.

The presets for pop placement are all wrong (other than "City Growth") so if the AI is using those it's probably also not specialising cities as much as it could be.


Tactical Combat - really good, not perfect. I think each unit is doing what it should, and AI targets injured units to finish them off, but there are some dumb moves like leaving fortifications or suicide attacks against melee when defending would have been better. Also co-ordination and placement on complex terrain is hard and not something AI is ever likely to be good at.


Strategic Combat - the AI seems to have good spacing of its forces so they have reinforcements, retreats when it should, and targets isolated units. However, its pathing becomes predictable and it will send lone stacks, so it gets possible to lay ambushes, and it sometimes does suicidal attacks for no reason.

 

[Krajzen]

I have some worries and suspicions about AI function on the macro geopolitical and strategic scale, such as

- #2 empire having almost no army when I attack it (despite not being in any other war)
- ...and not building it as quickly as possible after I have attacked
- ...and not recalling that one besieging army it actually has to save itself (but that's tied to the bug? of AIs sometimes falling into infinite sieges which literally never end, which is especially infuriating when then block your movements or catch your armies which can't move then)
- ...and not calling its two allies (!) to help, apparently?
- ...and AI empires in general not being agressive and expansive enough IMO. I mean, they do declare wars and capture stuff, but they are unable to 'finish the job' as in: fully conquer major factions and establish scary hegemony, they mostly get like one city and stop at that.
- What is also worrying is, I am not sure if they are capable of at least negotiating one territory from a peace deal, they seem to only get major cities - and there are very few of them so the map barely changes through the game.
- They could also be more reactive or opportunistic: I had no military for two eras and nobody came to exploit this fact.