2.8.1 (2190) Beta AI Feedback Megathread

  • We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Status
Not open for further replies.
War in heaven also is long overdue for an overhaul. The conditions for the ai for joining an afe side are basically, that their fleet strength is below 80k and that they are not part of a federation. Only bad that the time of when the war in heaven can break out was moved from midgame to early endgame (which is year 2400 by default). By the time the wih breaks out pretty much every ai empire is already part of a fed and most of them have over 80k in fleetpower. Due to this both afe always end up alone with estimated 20 empires (on huge maps) as enemies as part of the league. Reaching worldpeace way before the actual endgamecrisis spawns.
 
  • 1Like
Reactions:
War in heaven also is long overdue for an overhaul. The conditions for the ai for joining an afe side are basically, that their fleet strength is below 80k and that they are not part of a federation. Only bad that the time of when the war in heaven can break out was moved from midgame to early endgame (which is year 2400 by default). By the time the wih breaks out pretty much every ai empire is already part of a fed and most of them have over 80k in fleetpower. Due to this both afe always end up alone with estimated 20 empires (on huge maps) as enemies as part of the league. Reaching worldpeace way before the actual endgamecrisis spawns.

Normally you would expect a dynamically scaling value based on a function that takes several factors into account for this decision (AFE strength and proximity, (average) strength of other empires alive, ongoing crisis, etc.). But ofcourse something like this isnt really possible with the scripting system in stellaris, or if it is, its extremely clumsy to implement. I hope future titles have a better modding language.
 
Mid game cricis are incredibly weak. First of all they are not affected by the crisis strength slider or the ai difficulty (afe are an exemption, but their spawn fleets themselves also are not affected). The only way currently to make them bigger is increasing the galaxy size and even then they fail. After the 2.6 military overhaul the max crisis strength was increased from 5x to 25x because the crisis fleets suddenly ended up being too small. The mid game crisis types did not receive such a buff.
War in heaven also is long overdue for an overhaul. The conditions for the ai for joining an afe side are basically, that their fleet strength is below 80k and that they are not part of a federation. Only bad that the time of when the war in heaven can break out was moved from midgame to early endgame (which is year 2400 by default). By the time the wih breaks out pretty much every ai empire is already part of a fed and most of them have over 80k in fleetpower. Due to this both afe always end up alone with estimated 20 empires (on huge maps) as enemies as part of the league. Reaching worldpeace way before the actual endgamecrisis spawns.

Both of these are only an issue due to your own settings. Nothing needs to be changed about the mid-game or war in heaven, the default dates just need to be shifted earlier. At midgame 2250 and endgame 2325, both of these crises are well balanced. And nothing is preventing you from just changing the dates to those.
 
  • 3Like
Reactions:
Additional Update from my observations since my last post:

- AI still doesn't build enough shipyards and titan factories, though it improved slightly since the latest beta update. However, I would call the amount of shipyards being build still mediocre!! Please fix that for once and all!

- Also address to prioritize the AI to build megastructures as soon as possible, like the strategic command center and mega shipyard.

- Gestalt AI (especially machine empires) is not building enough unity buildings in my opinion. Machine empires were always at the lowest end of total unity production in my observed matches, followed by hive.

- In my playthroughs when the gray goo event happend the tempest was always way too strong for the AI empires resulting in that the tempest purged 80% of the galaxy (10 AI empires). Only two AI empires (gestalt) were strong enough to defend off the tempest (Difficulty: Commodore with scaling enabled and 2x crisis strength on a 1750 starsystem map)

- When machine uprising happens the machines were way too strong for the common AI. Some uprising machine empires spawned with 1,3M fleet power while their main AI species had a fleet power of around 300k'ish. Something seems to be generally weird with the scaling and difficulty and in comparison with the AI fleet strength. I sense a pattern here and I think one of the reasons is that the AI is not building enough shipyards to provide sufficient build capacities (for building ships and upgrading them)...

- btw I've never seen the AI to build juggernauts in my playthroughs or matches that I've observed.

- Prethoryn Scourge is doing nothing when spawned, but I think that was already mentioned here.
 
Strange - for me the prethoryn were pretty active (despite a lack of constructors, which caused their territory turning into a cheese as they cleansed more territory faster than they were able to built bases).

Also the grey tempest in my opinion highly depends on luck and the tech speed settings in the options. The higher the tech speed, the faster the ai rips open the cluster and the fewer it is capable to fight the tempest. Also a low ai difficulty is not so healthy for the ai fighting capabilities.

So far never seen the machine uprising getting even 1M on fleetpower - but that probably depends on the type of machine empire which spawns. It dies anyway - the feds almost everytime clean it.
 
Strange - for me the prethoryn were pretty active (despite a lack of constructors, which caused their territory turning into a cheese as they cleansed more territory faster than they were able to built bases).

Well, the prethorian spawned at the very top of the map where absolutely no AI empire had territory. Maybe an explanation for that is that if the crisis AI can't find an enemy in certain range of star-systems it will just idle and do nothing?! As I mentioned for that match, the map was very big and almost all empires got purged by the tempest resulting in alot of systems unclaimed and the remaining AI empires were probably not sitting close enough to the scourge.

Also the grey tempest in my opinion highly depends on luck and the tech speed settings in the options. The higher the tech speed, the faster the ai rips open the cluster and the fewer it is capable to fight the tempest. Also a low ai difficulty is not so healthy for the ai fighting capabilities.

This seems to make sense and could be the reason since I've set the tech speed like to the second, or third fastest setting.
On the other side this also means that the AI is not properly scaling its techs and fleets, imho.

So far never seen the machine uprising getting even 1M on fleetpower - but that probably depends on the type of machine empire which spawns. It dies anyway - the feds almost everytime clean it.

Something I forgot to mention was that the fleet strength changed sometimes from 1,3M to like 400k and then back again. I've no idea what was causing that.
 
I wrote this at work today. I hope it makes sense but this is how I would 'like' the Sector AI to work as well as a couple of my thoughts.

You need to decide if Sector automation / Planet Automation is an accessibility option for new players and should function reasonably well and be a player aid that can be relied upon and not mess up, it doesnt need to be optimal or work strictly to a "meta" (I hate that word) but it needs to be of a reasonable standard.

If it's not and is just a function to bring down micromanagement of planets, then you seriously need to bring it to a even higher standard of function because people will expect that much more.

I think the below is a good base line to go off and isn't horribly complex.


Sector AI feedback :-

What I think it should do (high level).

Balanced Build slots

The core (primary) sector should default to Balanced but be changeable (With consideration that the first planet of the core sector, the initial Capitol you cannot specify a designation anyway)

Additional planets in this sector should also try to get a healthy start with Gene clinics, Robot Assembly and buildings to counter low happiness as their priority.

And this should apply for any further Balanced sectors.


Then prioritise the buildings slots based on the following

1) Is the empire at/over the limit of Empire Sprawl? Yes? Build Admin offices

2) If ^ No, Is the food/mineral/CG production negative or close to for two repeated months for the sector? Yes? If there already more than 3 districts of the value thats low(energy/min/food) build the speciality bonus EC/min/food building. If CG is low, look at using a slot for creating more. (Limited to 2 per planet, on a Balanced setting)

3) If Research output in sector is below <some calculated value that grows on 'current year'> (maybe a average derived from the galaxy, per sector everywhere) and there is not more than 2 existing Research buildings, build a research building.

3) Is there a surplus of Minerals (e.g. 200+ and above conditions are not met e.g. No), build alloy foundries upto a limit of 3 per planet (It is a balanced setting remember)

Remaining slots to be used by other buildings including speciality resources.

District usage

[Balanced]Sector AI should correctly allocate a planet if it is to be energy/min/food focussed based on number of slots, whichever has the max (or if there are two very high maxes), use them and ensure the building slot used matches the "speciality". (e.g. Mineral Purification plants). If the planet itself is "roughly balanced" in max district distribution, then no priority for any type.

---

Unlike Balanced, these 3 "specialisations" should focus on 'specific' building slots until the need is 'met' then return to normal buildings to improve growth of the planet and increase pop growth/robot build speeds.

[Production] Sector AI setting is bogusly named / described and if it is, as described - its open to be misinterpreted. Since districts are not at all exclusive to a planet being "Manufacturing" or Balanced. A planet with significant alloy manufacturing can also be a planet that has 12 Mining districts and produce a large amount.

You could opt to simply remove this as an option, or, just better explain that it will prioritise buildings that enhance production along with districts, before defaulting to more normal "balanced building.

[Research] Sector AI setting should build Research buildings as long as it does not put the sector at a resource deficit, again it should use a baseline galaxy average on sector values and unlike the Balanced setting, aim to remain X above that value. Whilst maintaining than use a 'Balanced' build queue.

[Manufacturing] Sector AI setting simply needs to maximise the production of Alloys, followed by lesser priority items like CG. This should look at the Sectors +/- of Minerals and stop if it will plunge the Sector into a negative monthly value.
CG buildings should be limited to e.g. 3 per planet even with this specialisation whilst maintaining a healthy % EC surplus per month. When not building either, build normal 'Balanced' buildings.

When not building these buildings, create speciality resource buildings as the next priority (that are naturally occurring), before then creating a surplus of the manufactured type buildings.



^^ That would be the template I'd look to work with. And then when dealing with unique empires such as Lithoids/Machines, they need to deviate from this and exclude certain checks/buildings.
 
Last edited:
  • 1
Reactions:
2 very recent testing games on the latest beta saw some real inconsistent behavior with swarm.

Game 1 saw the swarm behave pretty aggressively but slowly. They didn't really back-fill their territory but they actually did stuff and infested planets. They seemed sluggish though, long periods of time when their colony ships just sat around and their armies did nothing while their fleets had full control over multiple systems. Their fleet AI was good and aggressive, they just didn't use their construction ships at all and were very very slow with their infesting ships.

Game 2 Swarm arrived in empty space, or at least space with no colonies. They were pretty aggressive with their fleets but they took no territory and did absolutely nothing else. With no infested planets they quickly petered out. They got into similar situations where their fleets absolutely controlled multiple systems with colonies, but their colony ships never infested anything, never moved. Troop ships also never moved.
 
  • 1
Reactions:
This fleet was going to attack the station until 3rd Star Order entered system - then it froze in place.
 

Attachments

  • ironman.sav
    2,8 MB · Views: 0
  • 20201118235335_1.jpg
    20201118235335_1.jpg
    284 KB · Views: 0
  • 1
Reactions:
Additional Update from my observations since my last post:

- AI still doesn't build enough shipyards and titan factories, though it improved slightly since the latest beta update. However, I would call the amount of shipyards being build still mediocre!! Please fix that for once and all!

- Also address to prioritize the AI to build megastructures as soon as possible, like the strategic command center and mega shipyard.

- Gestalt AI (especially machine empires) is not building enough unity buildings in my opinion. Machine empires were always at the lowest end of total unity production in my observed matches, followed by hive.

- In my playthroughs when the gray goo event happend the tempest was always way too strong for the AI empires resulting in that the tempest purged 80% of the galaxy (10 AI empires). Only two AI empires (gestalt) were strong enough to defend off the tempest (Difficulty: Commodore with scaling enabled and 2x crisis strength on a 1750 starsystem map)

- When machine uprising happens the machines were way too strong for the common AI. Some uprising machine empires spawned with 1,3M fleet power while their main AI species had a fleet power of around 300k'ish. Something seems to be generally weird with the scaling and difficulty and in comparison with the AI fleet strength. I sense a pattern here and I think one of the reasons is that the AI is not building enough shipyards to provide sufficient build capacities (for building ships and upgrading them)...

- btw I've never seen the AI to build juggernauts in my playthroughs or matches that I've observed.

- Prethoryn Scourge is doing nothing when spawned, but I think that was already mentioned here.

Different point of view: I had no less than 4 machine uprisings happen to AI empires in my latest game, and all of them were handily put down by the original AI. I feel like it might be your mod making the AIs too weak compared to the crises, or maybe Scaling Difficulty is just that broken. I recommend never using Scaling Difficulty regardless, because it starts the AI out extremely weak. That will color what you're seeing the AI doing. The AI did actually start building megastructures around 2400 in my game as well, although 2400 was also the victory date and one of the human players had 10+ megastructures by then, so they should definitely start much earlier.
 
Different point of view: I had no less than 4 machine uprisings happen to AI empires in my latest game, and all of them were handily put down by the original AI. I feel like it might be your mod making the AIs too weak compared to the crises, or maybe Scaling Difficulty is just that broken. I recommend never using Scaling Difficulty regardless, because it starts the AI out extremely weak. That will color what you're seeing the AI doing. The AI did actually start building megastructures around 2400 in my game as well, although 2400 was also the victory date and one of the human players had 10+ megastructures by then, so they should definitely start much earlier.

Ok, I will give that a try and let the game run without scaling difficulty. Maybe that's truly the reason why the AI feels so "borked".
 
Maybe it would be good, if the player could set economic goals for the sector ai during game like it is possible with the "Improved planet and sector automation" mod? This would give players more control for their specific needs, because no matter how good the sector ai becomes, different players will always complain, that they not get what they want. This would give players additional control without getting into microing.
 
  • 6Like
Reactions:
I created a vassal out of a sector in freshly conquered territory and gave the AI 2000 units of all non special resources. The AI then builds 4 colony ships despite having only having access to two planets within its borders. I would've expected it to build a construction or science ship or some military vessels after the first two colony ships.

Screenshot of the Vassal and Savegame is attached.
 

Attachments

  • 2020_11_19_2.png
    2020_11_19_2.png
    3,1 MB · Views: 0
  • ironman.sav
    2,2 MB · Views: 0
AI insist building agriculture district while there is a 80+ food surplus and the energy income is a one digit figure.
 

Attachments

  • 2257.01.19.sav
    2,4 MB · Views: 0
  • 2
Reactions:
War in heaven also is long overdue for an overhaul. The conditions for the ai for joining an afe side are basically, that their fleet strength is below 80k and that they are not part of a federation. Only bad that the time of when the war in heaven can break out was moved from midgame to early endgame (which is year 2400 by default). By the time the wih breaks out pretty much every ai empire is already part of a fed and most of them have over 80k in fleetpower. Due to this both afe always end up alone with estimated 20 empires (on huge maps) as enemies as part of the league. Reaching worldpeace way before the actual endgamecrisis spawns.

Fill out a bug report. This is off topic.

View attachment 655863
As I posted before, AI target planning is counterintuitive when there are hyperlane inhibitors...
More details required, especially since I can see you're not using debug_ai, and obviously haven't read the OP.
This fleet was going to attack the station until 3rd Star Order entered system - then it froze in place.
AI insist building agriculture district while there is a 80+ food surplus and the energy income is a one digit figure.

More information required. A one-line response isn't useful feedback here.

Maybe it would be good, if the player could set economic goals for the sector ai during game like it is possible with the "Improved planet and sector automation" mod? This would give players more control for their specific needs, because no matter how good the sector ai becomes, different players will always complain, that they not get what they want. This would give players additional control without getting into microing.

This isn't the place for out of scope conversations. We're not looking at reinventing the wheel here, simply looking for feedback on the current sector AI. :) Suggestions like this belong in the suggestions forum, thanks!
 
Status
Not open for further replies.