• 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.

Stellaris Dev Diary #21 - Administrative Sectors

Hi again folks!

Today I am going to talk about one of the great pitfalls of strategy game design; dull micromanagement. That is, features which require too much player attention. The trick, of course, is determining how much is “too much”, but it’s useful to consider how central the feature is to the core gameplay, how well it scales between small and large states, and how repetitive it gets with time.

In Stellaris, one feature which risked causing bad micromanagement was the planetary tile system; assigning Pops to tiles and deciding which buildings should go where. It is a fairly central feature and it is fun to use… but if you had to worry about 20, 50 or more planets, it would scale poorly. The obvious solution to this type of scaling issue is automation; you can let the AI handle it for you. This is indeed what we did in Stellaris, but not in a “traditional” fashion... Instead, we opted for something a little bit more akin to the vassals in Crusader Kings through something we call Administrative Sectors.

stellaris_dev_diary_21_02_20160215_edit_sectors.jpg


A Sector is an administrative region under the control of a Sector Governor. You can control a few planets directly (your “core worlds”), but once you go past the limit, you will start suffering penalties to your Influence as well as Empire-wide income. The exact limit for how many planets you can control directly depends on various factors, like your government type and technologies, but, as with the “Demesne Limit” in Crusader Kings II, it will never be a huge number. At this point, it is best to start dividing your territory into Sectors. You can decide the Sector capital and which planets should belong to it (but they must all be connected to the capital, i.e. form one cohesive sub-region.) You are also allowed to name your Sectors, for fun.

Unlike proper Vassals, Sectors remain an integrated part of your Empire, but they will handle development of planets and the construction of mining stations within their region for you. You can give them a focus (Industry, Research, etc), an infusion of Minerals or Energy Credits to help them along, and decide if you want to tax them for Minerals and Energy Credits. Sectors do not possess any military fleets of their own, nor do they perform research (they have access to the same technologies you do, and their research output is all given to you.)

stellaris_dev_diary_21_01_20160215_sectors_list.jpg


While Sectors and Sector Governors cannot demand more autonomy, or directly rise up in revolt (things I’d love to explore in an expansion), over time their population tends to diverge ideologically from that of the regime, and create their own identity. Like-minded Pops will tend to migrate there if allowed to. In the same way, aliens of the same species will also tend to coalesce in the same Sectors. Thus, when Factions form, they will often tend to have their main seat of power in a specific Sector. And Factions can demand autonomy and achieve independence. However, this is something that warrants its own dev diary...

That’s all he wrote folks. This time. Next week, I plan to talk about Alliances and Federations!
 
Last edited by a moderator:
  • 241
  • 70
  • 7
Reactions:
Considering that those rules are basically how I would specialize a planet, I'd certainly hope the AI follows them as well.

To some extent. I wouldn't want even a focused science sector to post only just sufficient food and science buildings. Needs extra food to grow POPs and minerals/energy to build other science buildings.
 
Yeah, there are exceptions: you need enough farms for max population, so if you have a really food-poor planet you might have to build extra farms. Still, the exceptions are simple enough to program, and self-evident enough for a human playing through it.
 
The imo sensible way to do it would be the following:

If a tile has only one resource, the AI places the appropriate building.
If a tile has the choice between 2 resources, the AI choses your specialisation.
If a tile has no resource, that AI places your specialisation.

This maximises resource output while still giving a boost to whatever you told it to do.
Tiles have adjacency bonuses though, so if the ai is good enough to handle this, im all for it
 
Tiles have adjacency bonuses though, so if the ai is good enough to handle this, im all for it

It's an optimisation problem. Computers are very, very good at optimisation problems, to the point where the AI of some games has needed to be dumbed down in order to give the human a fighting chance. I'm happy to let the AI handle it.
 
  • 1
Reactions:
It's an optimisation problem. Computers are very, very good at optimisation problems, to the point where the AI of some games has needed to be dumbed down in order to give the human a fighting chance. I'm happy to let the AI handle it.

an interesting point, I'm always been very interested in AI development: can you please give some examples of games in which the AI was so good that the developers was forced to dumbed down it ?
 
an interesting point, I'm always been very interested in AI development: can you please give some examples of games in which the AI was so good that the developers was forced to dumbed down it ?

NPC aiming in most FPS games is a good example of this. It's really easy to make the AI-controlled gunmen score hits almost all the time; aimbots have been a thing for many years now. Making the AI miss in believable ways is much harder.

I'm not an expert but I believe that timing in sports games is also an example of this.
 
Last edited:
Sure, as long as it knows what to optimise. Optimisation isnt always a matter of pure efficiency.
Exactly. The optimiser is... well, I was going to say "Only as good as the programmer", but it's actually a lot LESS good than the programmer because the computer is fucking stupid. Its only advantage(?) is that it is stupid a lot faster than the programmer is.
 
an interesting point, I'm always been very interested in AI development: can you please give some examples of games in which the AI was so good that the developers was forced to dumbed down it ?

i can;t give any specific examples, but a proper AI can be exceptioanlly good at managing a large number of units. let's take your generic RTS - an AI could always perfectly manage all his production buildings, make sure workers are on the right recources in the right numbers, manage patrols, reseed farms ect. whilst a human player needs to struggle to keep up with that effieciency.

basicly, the less complex tought required for something, the better the AI is at it. the larger the scale, the better the AI is at it(it's simply impossible for a human player to give indiviual commands to a few hundred units in a short timespan). to keep humans and AI on the same playing field, you need to force the AI to do only a certain number of actions each second or stuff liek this can become problematic. AI has perfect memory and perfect calculations - the human player does not.

for example, in LoL certain intermediate bots used to be terryfying because they could perfectly hit and chain abilities and the like, making champions that relied on that really scary(;_; cassiopeia bot) to the point they played the champion better than most of the playerbase could, and were later dumbed down to miss more abilities.
 
NPC aiming in most FPS games is a good example of this. It's really easy to make the AI-controlled gunmen score hits almost all the time; aimbots have been a thing for many years now. Making the AI miss in believable ways is much harder.

I'm not an expert but I believe that timing in sports games is also an example of this.

sorry, but since we are on the Stellaris - Paradox forum for me it was obvious that we were speaking of wargames or strategy games.

my original question would have been:

an interesting point, I'm always been very interested in AI development: can you please give some examples of strategy games or wargames in which the AI was so good that the developers was forced to dumbed down it ?
 
i can;t give any specific examples, but a proper AI can be exceptioanlly good at managing a large number of units. let's take your generic RTS - an AI could always perfectly manage all his production buildings, make sure workers are on the right recources in the right numbers, manage patrols, reseed farms ect. whilst a human player needs to struggle to keep up with that effieciency.

basicly, the less complex tought required for something, the better the AI is at it. the larger the scale, the better the AI is at it(it's simply impossible for a human player to give indiviual commands to a few hundred units in a short timespan). to keep humans and AI on the same playing field, you need to force the AI to do only a certain number of actions each second or stuff liek this can become problematic. AI has perfect memory and perfect calculations - the human player does not.

for example, in LoL certain intermediate bots used to be terryfying because they could perfectly hit and chain abilities and the like, making champions that relied on that really scary(;_; cassiopeia bot) to the point they played the champion better than most of the playerbase could, and were later dumbed down to miss more abilities.

I agree that the AI can do their things better than humans giving a short timespan - but if you pause your RTS game anytime you need and give the necesssary orders I doubt that this can really happen.

and in turn based strategy game or wargame I have still to see an efficient and competent AI than can do their things better than humans.

I would like to meet such a competent AI player that resemble to me the dreadful SkyNet, but in fact I'm scared of such a possibility.. o_O
 
I agree that the AI can do their things better than humans giving a short timespan - but if you pause your RTS game anytime you need and give the necesssary orders I doubt that this can really happen.

and in turn based strategy game or wargame I have still to see an efficient and competent AI than can do their things better than humans.

I would like to meet such a competent AI player that resemble to me the dreadful SkyNet, but in fact I'm scared of such a possibility.. o_O

that's ignoring AI strengths in favor of human strengths. we can figure out the best way to do things on ourselves, while the AI is only as good as it's specifically told to do. what it's actually good at compared to humans is not forgetting a single thing(it can keep track of everything at once), and being able to do everything at once. given infinite time, an competetent human should be able to win 100% of the time, because of the simpel reason that it's right now literally impossible to do better than the human brain. we haven't even come close to the amount of processing power needed to match that.

it can only beat us on speed, memory and precision - on these things, a fully functioning proper AI WILL beat every single human being given a long enough time. they're simply better at that than we are, if only because they don't have to deal with the same input lag we have to deal with. hence it'd do insanely well in a proper RTS you can't give orders while paused, and even in htose he can keep ordering units every single tick whilst a human would still be limited to reaction speed. for example, high level starccraft bots could using general tactics end up pretty high on the ladder, because of what they're good at. yet even the worst player can stomp them if they use some weird strategy the bots can't deal with.

an player is only as good as he plays, and an AI is only as good as it's coded after all though, but based on pure 'skill' AI's or even software in general are not even slightly better than humans. they're good at not making mistakes within the ruleset you give them and doing stuff really really quickly. and almsot all of them have non-standard ways to exploit them ebcause it falls outside of their ruleset - think CKII baiting stacks to attack you with tons of troops in the sea, or EUIV blockading shenanigans, or selling all your stuff to the AI in civ for a lump sum then decalring war on them to get everythign back. humans can rely on common sense to not make a lot of obvious mistakes, whilst an AI is only as good as you teach him to be and is limited by processing power.
 
  • 1
Reactions:
I agree that the AI can do their things better than humans giving a short timespan - but if you pause your RTS game anytime you need and give the necesssary orders I doubt that this can really happen.

and in turn based strategy game or wargame I have still to see an efficient and competent AI than can do their things better than humans.

I would like to meet such a competent AI player that resemble to me the dreadful SkyNet, but in fact I'm scared of such a possibility.. o_O

The ai in AGEOD's American Civil War was pretty good, especially if you turned on the extra resources option but that's turn-based so not really comparable to PDS style games.
 
The ai in AGEOD's American Civil War was pretty good, especially if you turned on the extra resources option but that's turn-based so not really comparable to PDS style games.

I agree, the AGEOD series wargames AI is not bad: I played both "Birth of America" and "The American Civil War" and that French guys know how to program a good-enough AI opponent.

but turn on the extra-resources option in fact is not so much different than giving the AI the permission to slightly cheat.. o_O

giving the advantage of the human brain vs AI, it can be anyway acceptable for me.

but I would strongly prefer to find a stimulating AI opponent that operate with the same playing conditions.

speaking of turn based strategy game vs Paradox style grand-strategy games I do not find them so much different - in fact I think that at their hearts the Paradox games are turn based, since you can pause them at will.

or at least this is the style I use to play them, so for me they not differ so much from turn based strategy games or wargames.

ps: I do not play the Paradox games or any other pausable RTS game vs human players but only against the AI.
against human players I play only turn based strategy games or wargames both WEGO or IGO-UGO.
 
Last edited:
Will this anger the Sector's Governor or cause an increased revolt risk in that Sector, like it would in CKII?

It's balanced by a resource, influence. But it's not medieval feudalism, so the player seems to have more control over their sectors.

Revolts would be caused by pops migrating to sectors, where they can group themselves into powerful rebellious factions with rebel leaders.
 
  • 1
Reactions:
What confuses me the most is they put in a mechanic like adjacency tile planet building - the epitome of strategy micromanagement gameplay - and then something like this and say that they are trying to avoid too much micromanagement.
 
  • 2
Reactions:
What confuses me the most is they put in a mechanic like adjacency tile planet building - the epitome of strategy micromanagement gameplay - and then something like this and say that they are trying to avoid too much micromanagement.
How is this confusing? Players want to micromanage/control a few important things, just not necessarily repeat the same task x20. See: every space 4x game once the player's empire balloons too large.
 
  • 1
Reactions:
But the adjacency is only for the capital. In your regular 4x5 (i.e. 20 tiles) planet, only ONE building will give adjacency bonus, and this one building will affect, AT MOST, 4 tiles.
 
  • 1
Reactions: