Ways to improve ship design diversity for ai

[forfor]

Right now every ai autodesign is spamming autocannons. It really sucks in the sense that once you know everything is going to be autocannons it becomes very easy to counter, and at the same time it encourages the use of monofleets, which was what the recent ship redesign thing was supposed to discourage. I'm curious what people's solutions to this would be? Or do you think it's fine as is? my personal idea is to have certain components and builds get a bit of extra weight in the autodesigner based on the empire ethics. For instance, spiritualist empires could get weighted toward carrier builds, pacifists could be more weighted toward artillery builds, etc. Of course, the ai should probably also get some weight toward selecting the relevant researches for those weapon types when researching.

This isn't a perfect solution, but it would at least encourage some diversity, and it would give some thematic rp to the AI.

 

[Dragatus]

I'm just going to repeat a portion of what I posted in your other thread.

I think small ships being fragile is pretty much working as intended. What isn't working as intended is the AI ship designer. This is because some time ago someone at PDX came up with the idea that the AI should use how much power a weapon uses as a proxy for how good the weapon is. And I think that was actually a pretty clever idea, it's just that the combat rebalance didn't take it into account properly and now the AI thinks that plasma and autocannons (which use abnormally high amounts of power) are high tech superweapons while swarmer missiles (which use very little power) are low tech trash. It's so bad that even AI empires that are supposed to favor explosive weapons will still rather use a plasma & autocannon combo instead of using swarmers. And it's the main reason why all late game AI designs look so similar.

The simplest and least disruptive fix is to standardize weapon power usage so that all weapons of the same size and tech tier use the same amount of power. I tested that out and can confirm that the AI starts using more diverse designs.

 

[Chuunibyou Imouto]

Personally, I would rather the AI pick weapons based on their empire type / personality / situation etc, rather than just using the auto ship design.

I think the combat rework introduced some great stuff, and the system is now there for the devs to easily build on. They added buttons to encourage the auto designer to focus on a few basic roles, so all they need to do now is add a bit more to where the AI will focus on Kinetics or Lasers or Missiles etc.

This would make Intel far more useful, and also make counter ship building a very important thing so you don't roll up to fight and not realize the enemy AI built all armor and missiles and eat your booty like the groceries because you weren't prepared for it.

Empire personalities in general could be used to do a lot more IMO. A hyper defensive focused pacifist that cranks out as many defense platforms and bastions as they can, or a hyper aggressive swarmer focused empire that just spams the crap out of disruptor corvettes etc would make the PvE ship meta far more interesting between games

 

[Xaelyn]

Right now every ai autodesign is spamming autocannons. It really sucks in the sense that once you know everything is going to be autocannons it becomes very easy to counter, and at the same time it encourages the use of monofleets, which was what the recent ship redesign thing was supposed to discourage. I'm curious what people's solutions to this would be? Or do you think it's fine as is? my personal idea is to have certain components and builds get a bit of extra weight in the autodesigner based on the empire ethics. For instance, spiritualist empires could get weighted toward carrier builds, pacifists could be more weighted toward artillery builds, etc. Of course, the ai should probably also get some weight toward selecting the relevant researches for those weapon types when researching.

This isn't a perfect solution, but it would at least encourage some diversity, and it would give some thematic rp to the AI.

Congrats, you have in invented

AI personalities - Stellaris Wiki

stellaris.paradoxwikis.com

They already do this, though maybe the weighting is a bit too low in the current patch.

 

[forfor]

Congrats, you have in invented

AI personalities - Stellaris Wiki

stellaris.paradoxwikis.com

They already do this, though maybe the weighting is a bit too low in the current patch.

Oh, I'm 1000 hours in and I literally never knew this was a thing, good to know.

 

[Ryika]

Thing is, why wouldn't the AI use auto cannons?

If a weapon has up to three times(!) the tooltip dps of other weapons in that slot with only small downsides that don't compensate in any way (and with a universally useful +25% against armor modifier on top), using lasers or mass drivers would be a pretty bad idea. The AI does not understand disruptors, so it probably should not use them either unless the devs can get the AI to use them well, and missiles on small ships are in kind of a weird spot in general.

Auto cannons are as close as the "objectively correct" choice as you can get for the AI (and the s-weapon slot in general), and if the goal is to make the AI use other weapons too, then a proper rebalancing of that weapon is pretty much necessary as well.

 

[SirBlackAxe]

Autocannon fleet power calculation might be the most important issue to fix currently

I understand development can't happen overnight and different people will have different priorities, but I hope to argue here why the current state of autocannon is influencing major aspects of the game negatively and need to be a priority. For...

forum.paradoxplaza.com

It's hard to offer concrete solutions to this when the AI weight calculations for ship components are so opaque. I wish you could check the values with debugtooltip.

 

[forfor]

Thing is, why wouldn't the AI use auto cannons?

If a weapon has up to three times(!) the tooltip dps of other weapons in that slot with only small downsides that don't compensate in any way (and with a universally useful +25% against armor modifier on top), using lasers or mass drivers would be a pretty bad idea. The AI does not understand disruptors, so it probably should not use them either unless the devs can get the AI to use them well, and missiles on small ships are in kind of a weird spot in general.

Auto cannons are as close as the "objectively correct" choice as you can get for the AI (and the s-weapon slot in general), and if the goal is to make the AI use other weapons too, then a proper rebalancing of that weapon is pretty much necessary as well.

That sounds good in theory, but in practice it means all I have to do is build a bunch of ships with max range weapons and armor to 100:0 the ai

 

[Ryika]

That sounds good in theory, but in practice it means all I have to do is build a bunch of ships with max range weapons and armor to 100:0 the ai

Auto Cannons are so strong, if you pit 2 fleets of corvettes against each other, both with maximum armor, one with only lasers (which has +50% against armor), and one with only auto cannons (which has -75%(!) against armor), then in most situations the outcome is that cannons will STILL win the fight, albeit not by much, or lose only very slightly. They literally outperform or keep up with every other weapon in the s slot even if countered perfectly.

And long-range weapons just counter corvettes in general, don't think that has much to do with the weapon choice.

 

[forfor]

But that only really applies early game when you're using small ships. By late game, you get a lot more value from artillery ship alpha strikes, torpedoes, repeatable techs (I always seem to have way more physics repeatables than the others for whatever reason) and larger ship designs that can fit a lot more armor slots.

 

[CocoCincinnati]

There's one very easy and very obvious solution, make autocannons small weapons only again. That way, at least the AI won't use M and L slots for them.

Edit: This would also solve the problem of starbases with too many range 35 weapons on them.

 

[Ryika]

But that only really applies early game when you're using small ships. By late game, you get a lot more value from artillery ship alpha strikes, torpedoes, repeatable techs (I always seem to have way more physics repeatables than the others for whatever reason) and larger ship designs that can fit a lot more armor slots.

But what does that have to do with the weapon selection of the s weapon slot? Auto Cannons are still the best pick for it, even if the s weapon slot itself is not a good pick.

 

[DamnedLackOfTropicalFruit]

Auto Cannons are so strong, if you pit 2 fleets of corvettes against each other, both with maximum armor, one with only lasers (which has +50% against armor), and one with only auto cannons (which has -75%(!) against armor), then in most situations the outcome is that cannons will STILL win the fight, albeit not by much, or lose only very slightly. They literally outperform or keep up with every other weapon in the s slot even if countered perfectly.

And long-range weapons just counter corvettes in general, don't think that has much to do with the weapon choice.

This isn't true at all. Lasers don't lose, they win with less than 20% casualties!

Even if you're lucky enough to get nanite autocannons, simple gamma lasers still comfortably win, taking only 40% losses against an equal size fleet (less against an equal cost fleet). Autocannons get absolutely mauled in this comparison if you give the enemy corvettes a real brawler weapon, like plasma or PD.

Autocannons are fantastic, and are absolutely a key part of any brawler loadout, but they're not the "objectively correct" choice you paint them to be. Autocannons are ineffective against armor, extremely power hungry, and increasing the size makes them less slot efficient, much less power efficient, reduces their tracking, and doesn't increase range. S slot autocannons aren't the solution to everything, and M/L slots are just a trap.

I don't think this is a balance problem like you make it out to be - the AI favouring autocannons on corvettes would be fine if that was what the AI was actually doing. In reality, the AI uses nothing but autocannons on not just corvettes, but also larger ships and artillery.

 

[Zander]

Honestly, the takeaway I'm getting from this thread is "you should always use armor and only armor; shields are garbage", since the AI uses autocannons only.

 

[Ryika]

This isn't true at all. Lasers don't lose, they win with less than 20% casualties!
View attachment 946453

It's absolutely true for the majority of cases. I tested multiple tiers, and a variety of setups.
Only specifically in tier 5 does the fight become one-sided against Auto Cannons and a large part of that is the rising percentage of the health pool that consists of armor (plus passive tracking being very useful for laser), but again... this is them being hard-countered in terms of defense, and they still aren't wiped off the battlefield.

Autocannons are fantastic, and are absolutely a key part of any brawler loadout, but they're not the "objectively correct" choice you paint them to be. Autocannons are ineffective against armor, extremely power hungry, and increasing the size makes them less slot efficient, much less power efficient, reduces their tracking, and doesn't increase range. S slot autocannons aren't the solution to everything, and M/L slots are just a trap.

Wasn't talking about M/L slots, I specifically stated that I'm talking about S slots balance.

They're power hungry, sure, but the difference for 3 weapon slots is a maximum of 18 power on tier 5. That only matters in the sense that it takes away the option to run shields on your Corvettes.

And sure, they're weak against Armor, but they're still by far the best overall weapon for that slot. As a result, in a mixed fleet like the AI fields them, putting an anti-armor weapons into an s slot just diminishes the possible dps. On in other words, if you fill every s slot with an auto cannon, and put anti armor elsewhere, you'll get a much stronger overall package than if you put anti armor in your s slots. And yes, from my testing, that holds true on a real battlefield as well, even with the fact that longer-range weapons will shoot into shields first.

I don't think this is a balance problem like you make it out to be - the AI favouring autocannons on corvettes would be fine if that was what the AI was actually doing. In reality, the AI uses nothing but autocannons on not just corvettes, but also larger ships and artillery.

Didn't claim otherwise, I'm specifically talking about s slots on small ships.

/edit:
Just read through my first post again, and realized that I didn't actually specify that I'm talking about the s slot until the very last paragraph. I assume that's where the confusion's coming from - my bad.

 

[forfor]

I think the larger problem is that the conversation wasn't about small slots to begin with. If your design has small slots available, then autocannon spam makes some sense. The problem is that the ai is filling every slot of every size with nothing but autocannons, and sometimes a couple plasma launchers. This makes them very low range, and very easy to counter for a smart player.

 

[-Marauder-]

Honestly, the takeaway I'm getting from this thread is "you should always use armor and only armor; shields are garbage", since the AI uses autocannons only.

Kind of. Shields also got much weaker because so many weapons either pierce them entirely, or deal bonus damage.

On topic. Auto cannon damage numbers are so high, the Ai tends to use them. The game ignores all their drawbacks and completely overrates them both in how likely they're to be used as well as the fleet power it thinks ships armed with them have.

A fleet of Swarm Missile/Torpedo Cruisers will destroy a fleet of Cruisers/Destroyers/Corvettes armed with Auto Cannons without even getting the paint scratched. Despite the latter being shown as much stronger.

 

[Imp0815]

The system is inherently flawed.
Either you play rock-paper-scissors or you outnumber your enemy. One is somewhat micro intensive and requires some player engagement but its more of a To-Do list from a work environment:
-Get Intel
-Check what they use (Stone)
-Set all my Fleets to Paper!
-Win
With a typical build per empire type the first part is even cut off as player soon will know which AI empires will use which setups.
This will be the only thing we add. Now we know AI will always use autocannons/plasma. We would just add one extra step to a system that is already simple and bad to engage with.

The other way is just spamming ships until you outnumber your enemies. Then it does not really matter what the enemy is bringing to the table.

We lack mechanics that build upon the stats and numbers and give us some meaningful interactions with our fleets besides stacking them up and watch an entire empire worth of ships going up in flames in mere seconds.
Stats and composition should only make up a part on why and how battles are won and strategical elements need to be introduced to remove the point and kill interactions players currently have for warfare. It also needs to be slowed down so Combat can be actively observed and meaningful information can be extracted from it. This information can then be used to react and in turn the AI should be able to do the same.
But as of now its all happening in such a short notice that reacting in a state of war is just not feasible nor fun.

 

[Silesian Burd]

Just a heads up: the AI was already using only Kinetic Weapons before Orion. It took only DPS into account, and so the default Ship Designer would just give everything Autocannons and Kinetic Launchers, even in 3.5. If you had all tech, but didn't use the Ship Designer, your ships would be loaded with Kinetics and nothing else.

It's good to know that SP Stellaris was only made worse by Orion. And here I thought I could return to the game...

 

[-Marauder-]

Just a heads up: the AI was already using only Kinetic Weapons before Orion. It took only DPS into account, and so the default Ship Designer would just give everything Autocannons and Kinetic Launchers, even in 3.5. If you had all tech, but didn't use the Ship Designer, your ships would be loaded with Kinetics and nothing else.

It's good to know that SP Stellaris was only made worse by Orion. And here I thought I could return to the game...

The worst part is how it makes fleet power meaningless, with all the bandwagon issues that entails. You can either also go full Auto Cannons for maximum fleet power on paper, or suffer massive losses in diplomatic weight, when dealing with vassals, and so on.