Enemy procedural generation still a thing?

Phoenix Point
1

Are enemies procedurally generated?

If I remember well, the original concept was for enemies to be “put together” from different body parts when they are mutating. You would never know what kind of combination you would encounter - some would synergise better, some not. When the game stumbled on the variation whch cause you problems it would make more of them, until you adapt.

From videos I have seen I got an impression, that what we have seen so far are couple pre-determined builds in rotation, rather then full on generation - for example I haven’t seen a crabman using a gun and a shield.

If such change happened, I think it is understandable - designer can put together variations of crabman which would be fairly balanced, however I am not sure if “one type per lair” should be preserved in this case. I would rather see interesting mix of enemies in the same mission and have enemy base progression be expressed differently - be it in enemy numbers, or different enemy “builds” having different difficulty rating, with some appearing only when there is more then on lair operation in the area, or simply by deplying buffed versions of core builds.

If generation is still present but somehow muted (like some parts not appearing together to forbid OP enemies), what if a point system was introduced? Every body part would have a value, and PC would have limited amount of points when generating an enemy, so a more effective ones (like gun+shield) couldn’t be constructed by single base enemies. However, as enemy expands their presence in the area their point pool would increase allowing for a more deadly and more difficult to disable combinations.

2 Likes
2

We can only try to play the guessing game here. That’s very interesting and important quesion, and one of the weakest spots in game now. But I still confused to hear that question from you, who declared readiness to wait almost anything anytime and anyhow :slight_smile: Just because I feel us swaped roles and it’s you should answer me something like:
probably developers already made function of '‘dynamic spawn generation’ but switched that off for BB5 to keep privacy and finish their private balancing! We cannot guess what they exactly do, but I’m sure Julian know what he do, keeps promises, and his team keeps silence now just because they are too busy. Anyhow I wa ready to any result when prepaid and faithfully hope that team will will realize whole potential. You know, like this :slight_smile:

3

I think they are. It is just still most random.

Someone was lucky then not seeing this variant. :slight_smile: But I think that Arthrons (crabman) are quite well generated with their possible variants. I have seen less Triton variants than are possible. But well, they can vary just with all the different weapons, so their body parts are not so important anymore to provide great variety. :slight_smile:

there is one type per Nest. In Lair there are at least 2 variants. In Citadel at least 4.

I like this idea.

Currently aliens progress with “research” and all their mutations (for Arthron and Triton) will be swapped by upgraded body parts later on, but it is hard for me to say how quickly this jumps in and if it is inevitable (question without answer: will aliens go back to not upgraded versions to balance weaker player actions?).

4

By the way, if possible, can I ask to give me few examples of strategy games, where procedural generation of enemies was successfully realized? *I mean of course, ‘procedural generation of enemies’, actually based on your previous actions. Survivorship bias is my favourite .

5

@0nton I am against judging game’s state based on BBs. When it comes to discussing planned mechanics or it’s implimentation based on what was promised and comparing to what was presented in BB: I have quite the opposite stance. I am not complaning about how few variations seem to be in beta (which is the case) - I assume there will be more in 1.0. Just curious about the current implimentation and what we can expect to see in the future.

I can’t think of another game doing anything of that kind, no. Perhaps, MGSV? With enemies developing new bits and bobs to counter your tactics? Like puting on helmets preventing head shots, if you head shot too much? Though what PP was aiming to go for, was a random system, not directly responding to you, meaning you would never quite know what you will go against.

Which I like. My kind of rogue-likes are those which procedurally generate obstacles for you to deal with, rather then provide fairly static obstacles and vary up your toolset. PP seems to give you control over your development allowing you to shape the technological progress of your organization and tools at your disposal, while messing with the threat you are facing to keep things interesting. Conceptually, I dig this design a lot.

@Yokes Thanks, so my impressions were simply wrong. I was mostly interested in enemy generation and how it connects to overall world map progression. My impression must have been a result of few body parts in the rotation, and limited experience with the BB5. Of course, I did meant Nests - as someone who didn’t have hands-on experience I am fuzzy on much of the terminology.

That’s good, that means enemy variety can be fairly efficiently expanded - if not my main team, then it sounds like a potential playfield for modders, if such level of modding were supported. This is certainly one of the more intriguing aspects of PP to me, though as it is with experimental and ambitious mechanics, I have many plenty doubts regarding their practicality and fun factor. The concept has lots of potential, though.

6

@Wormerine As I said, announced procedural generation is very ambicious but complicated task. For having serious perpectives, it had to be presented already, at least partially, earlier than 3 month to release. Here are some points, why I think so:
*First, should we make a poll about aliens mutations: which are bugged/overpowered (like mist emitters, or grenade launchers aka equip crushers), more dangerous, or harmless, or useless?
*How it will be balanced with sametime rebalanced soldiers abilities?
*How AI will use random combinations of bodyparts?
*How will be automatically balanced different combinations of mutants in groups?
Too many blind spots, where we can find endless bugfield of extremes from exploits to overpowereds, randomly.
*Finally, we have no feedback about developers priorities.

So… We can only hope that developers plan to use stable and easy tested / fixed simple templates. Too late for difficult desicions I think.

7

if it is to follow original plan, the system which dictates mutations isn’t to be complicated at all. If I remember well it were to work as follows:

  1. aliens create a mutation of an alien constructing it from available body-parts at random
  2. alien is send against you - if it performs well, alien produces more of them. If it doesn’t, alien creates another random generation.

The idea was, that different body parts will be vulnerable to different strategies/items, so they weren’t supposed to be flat out better/worse - they would be better/worse depending on how you were playing, and hopefully, every once the would force you to change your play style. System were to be randomised to prevent “gaming” of the system.

The dangerous Bits here are:

  1. body part design - are they really varied and impactful enough to force change in tactics
  2. tools at our disposal - will the final game have enough variety in class builds and items/weapons to support unilinear and tactical adjustment to enemy evolution.

I am not terribly concerned about enemy/our weapon balance just yet. If after launch it would appear that some alien body parts are too useless or too powerful the can be buffed/nerfed.

If certain builds/equippement turns out to be an answer to all possible mutations it can be nerfed as well.

Those kind of thing can be well discerned with telemetry and player feedback. I am not sure if I would want Possible mutations/research t be spoiled before game’s release.

8

It’s not strategy, but Left 4 Dead 2 did a lot of good stuff with procedural generation, including enemies

9

Yeah, that’s pretty much my concern too. - It doesn’t matter how the AI is procedurally generated if you can just go shoot them all during turn 1.

10

Procedural generation in general is a bit different to what I’d originally thought it was going to be. When PP was originally announced I’d expect to have creatures evolve into one another, so for example you might initially face a horde or worms, but those worms would eventually become something else, some might become quicker, others would develop armour, maybe eventually the successful ones would become what we currently know as crabmen, and then those crabmen would evolve from there…

What we’ve got at the moment, feels like just a standard set of enemies, all be it with different load outs.

I agree also, that I’d rather see different variants of a given enemy type appearing on a map at the same time, facing x off all the same creature, isn’t as interesting as facing that creature with multiple varients IMHO.

12

Here is the point! How AI will understand what is perform well and what isn’t? I will show you two examples:

  1. Into the battle, almost all tritons with ‘Pithagoras’ were killed first and deal no damage in afterbattle statistics. What that means - they are weakest or most dangerous? So obviously… for human. But same as arthtrons without shield will be killed first. Does it means that these arthrons were most dangerous than without it?
  2. In afterbattle statistics AI can see, that more of the damage were dealt by claws in close range. Does it means that claws are most dangerous and should be generated in priority, or it was happened only because human gots fun by paralysing every alien, spiting of damage for his berserker?

I really hope that developers collect telemetry and feedback, and use it for making a handmade simple build templates. Because what we got now it’s just a bunch of cloned stamps with no inner tactics and no strategy. By the way, I cannot give any grades for aliens’ bodypart usefulness for last few weeks because all aliens spawns were already graded like ‘I feel no sense to give them chance for a long battle’. Who is the most interesting and dangerous opponent now? NJ squad with 3-4 armored heavies, 3-4 hidden snipers, and 3-4 assaults with interrupts, and 1 Armadillo - they only one who deserves personal approach here. And we can only believe that aliens will be changed as something new but same interesting.

Oh yes!.. From almost-animals to Lovecraftian monsters! But now… neatly drawn but cloned fully random loadouts :frowning:

Unfortunately, we cannot consider this example, and not because it is not a strategy, but because it is multiplayer. Do you understand why? S/L of course. Not because of a phenomenon and an opportunity in itself, but as a factor of diversity (not) affecting on procedural generation. In the end, for AI, everything that happened before the last boot is a blind spot, right? Accordingly, each loading increases the gap between the player and the procedurally trained AI. This is same as to compare the game in the mode of iron will and in the mode with saving games, even with twice as much difficulcy - you know what I mean?
On the other hand, if developers will find the way to dynamically increase AI strength depending on the amount of S/L produced by a human player, we run the risk of getting an invincible imba machine. Yes, we will definitely get it :slight_smile:

13

It wasn’t dynamic, but a relentless increase in difficulty until you eventually ran into an invincible opponent was pretty much every arcade game back in the 80s.

14

There’s also the question of how will the AI understand if you’re intentionally doing this? What's to stop us gaming the mutation system?

1 Like
15

You may Set aside about procedural mutations and consider this option: taking 2-3 superheavy units, 3-4 specialists and 4-6 melee units as a basic sample, give them a total of 10-12 sets of meaningful functions (rather than random!) through combinations of body parts. Something like this:

  • 3 chirons with acid worms, 4 armored arthrons with grenade launcher and 6 mobile tritons with shotguns - this team can supress armor;
  • 2 sirens with suppression of will, 4 newts of support with fog, 6 newts with paralyzing tentacles - capture team;
  • 3 armored arthrons with machine guns, 5 tritons with sniper rifles, 5 tritons with fog (if it will be dispersal after 2-3 turns) or shotguns - the initial team of snipers, etc.

And so on, there is still enough place for meaningful builds. I’m sure human will do it better then AI.
And then it’s already possible to endlessly increase the power of these builds, improving the functionality of body parts or better equipment, with keeping of main line. For example, increase armor until the player decides to change his shotguns to something more powerful or advanced. This way looks much more realistic but even it still requires incredible balancing.

16

I am pretty sure that from the beginning it was fairly transparent that mutation will be within each enemy type (remember various mutations of crabmen from fig timeline), rather then mutation from one spieces to another. I don’t really see a gameplay appeal of the latter on astrategic scale: that would essentially be “enemy tiers” like in XCOM1 and I prefer for enemies to stay relevant throughout the game.
And interesting idea might be enemy power evolution throughout one mission: so a small alien who becomes more and more dangerous as the mission progresses if not killed. A bit like a xenomorph. Perhaps feeding on civilians?

The more complex the system would be, the bigger chance it can be exploited, broken I am pretty sure, the idea wasnt to create a complex AI to analyse your tactics and adapt.

From what I imagined the system were to be system: if enemy manages to damage you or kill you, or fail your mission they would be a success.

As of now it is difficult to guess how such system would work, and you are right, it sounds like the most troublesome part of the system. Especially now, when loosing soldiers doesn’t seem common or viable I can’t guess how “effectiveness” could be estimated.

I don’t believe it was ever intention to intelligently analyse troopers performance (claws are good, let’s focus on claws). Rather rolling a dice again on mutation, once something tags current mutation as ineffective.

Ideally you want something like enemy AI from UFO: something really basic, but that makes alien seem to adapt by cosealing how things really work.

17

I agree to forget about complicated dreams. Although how cool it would be to see the transformation of the newt from a simple aquaman with a pistol to a six-armed mutant, chained in armor, with an acid bomb on a freshly tailed tail!
But let’s be realistic and try to guess about these two points: “rolling a dice again on mutation” and “enemy tiers” like in XCOM1"
Let’s start with the rotation of successful / unsuccessful mutations. Random switching is currently implemented. This mechanism at the moment looks completely random. In order for it to become interesting, absolutely different mutations should be ranged. However, by what principle a mutation utility scale can be implemented? What is more useful - agile legs or armored legs, if all enemies are slain and all soldiers survived? Okay, the game ran the full cycle of all possible mutations based on random - it’s time to choose the most successful build. How to do it? If will not use very unreliable parameters such as “damage done” or the “average lifespan in turns”, I see only a table of mutation with scores, manually determined by the developers.
Let’s suppose we have a table of body parts in conditional Mutant Points: Armor 10 = 2MP, Armor 20 = 4 MP, etc. Base pistol = 1MP, initial machine gun = 2MP, grenade launcher = 4MP and so on. At first, 8 points are assigned to the formation of the mutant, and further on increasing. Armor is growing, good. Damage is growing - great. But how should mutations with different functions be evaluated?
What about functional combinations of mutations? Probably we should again assume the basic patterns defined manually. For example, melee arms should be combined with agile legs, a chameleon ability, etc. And I still cannot guess how can be logically determined a moment to switch different kind of alien arms for example. Except one more level of templates - for alien squads, determined manually against different PPsquad builds. So, here is no place for any procedure generations on all steps from body parts to spawn combinations - all should be determined manually. Or I missed something?
Last step for dynamic is ‘alien tiers’ system. This system is simple, understandable and reliable, but unoriginal and has long been tested in different games - why has it not yet been applied in BB5?
*Oh, I actually missed the fact that we can get modified triton from the weakest (conditional 8 MP) to a terrible monster (more than 9000 MP) - however, I meant that this growth should be recorded in the mutation tables manually, I see no reason to call it procedural generation.

18

Not that complicated to increase and change 1) Mutation 2) Items 3) Level/Stats of aliens. Would be really cool, for some DLC or Mod to see excactly that (if not implemented already and we just saw a little placeholder).

The biggest problem is, as you already mentioned, how to determine what is the best way to mutate against your tactics.

19

Yes, the effectiveness of any mutation can be increased indefinitely according to the same principle that we increase the parameters of soldiers by simply issuing these ‘Mutant Points’ to the mutants. I just don’t see any reason to call it procedural generation if it is not based on some kind of AI self-training procedure. But this seems to be a matter of terminology :slight_smile:

20

It’s difficult to discuss how aliens could develop, without knowing how PP will develop.

The original pitch was that PP tech progression will be unlinear - we will develop tech with different utility, while not per-say better then each other. As such one one playthrough you might be initially better equipped to handle armored enemies, but not mobile DPS focused ones, or vice versa.

We haven’t seen such flexbility in BB5, so it is difficult to say if it is still something they still aim for. If such branching progression were implimented it makes sense to keep alien mutation random, to allow for catching the player off guard. Ideally, individual mutations would be strong against certain techs and build, while vonourable to others therefore not making them flat out better or worse overall.

There is a problem with tagging certain body parts as particular type (speedy legs with melee etc.) - then the mutation stops being what it is, but what we get is a set of pre-designed archetypes. Even if there will be couple different mutations which can be paired with the claw, what we still get are variations on melle-crabman, rather then mutation of crabman itself. Ideal situation would be if all limbs would add benefit for all possible weapons. So speedy legs could allow claw and granade alien to close the distance faster and attack, while they could also allow to jump up the ledges, and allow for better shooting angle for the gun. The more synergy between all parts of crabmen, the more crabmen variations we can meet and fight. But yes, it is something devs need to take a look at, and judge whenever random system generates adequate enemies on regular basis, or if it needs “God’s” intervention.

If certain combinations create flat out stronger mutations, it might be wise to indeed create a system, where certain items are impossible or unlikely to appear at the same time in early game. If synergy would be the biggest offender, then the more I am thinking about it, a point system might not cut it. If, lets say, a claw and speedy legs turn out to be super powerful, rating them both high would be impractical, as they in themselves aren’t the problem - it is their combination that is. However, if individual body parts turn out to be more powerful then others, then sure, they could be rated, with aliens being “put together” out of limited amount of points to spent. However, if such thing would happen, it might be smarter to buff/debuff things to be on more balanced, rather then restricting what can appear when and with what.

An increasing point system could be interesting if there were varius tiers of each mutation - so as the game goes on, aliens could start deploying enemies with superior version of a gun, claw, legs, armor etc. with amound of upgraded body parts increasing as time goes on. What part is upgrated could change inbetween mutations. But I don’t think it would work well for actual assembly of alien. Though it is difficult to say, without knowing what exactly will be on the offer.

21

I wouldn’t have a problem if a new mutation that was generated to defeat me, would force me to flee just because I can’t deal with it. I would have to change my gear, tactics and come back later or just loose the battle.
As example, there would be a mutation with massive armor and I don’t have anything to deal with it. Than I switch to armor shred/pierce weapons to defeat them.
As a result of it, they will use light amor, more mobile, melee and a high life mutation.

And so on.

Not so complicated, just a lot of variables for development.