Sector Discussion Thread

[emperor_kk]

Still the problem is not people complaining about sectors, but people complaining EVERYWHERE unrelated about "sectors will still be ****" as last seen in the "faction changes in 1.4 1.5 thread" (Pretty snarky post there by MementoMori, maybe Wiz' wording was kind of a direct response?). The real life equivalent would be people discontent in a restaurant with farfalle noodles always butting in in unrelated table conversations with "well the farfalle noodle is still sh*t!"

Also not all the people who make an account to this forum do that because they want to complain about sectors. Not even all the people who log in to the forums are complaining about the sectors. Not ever even all people who log in here to complain, complain about the sectors. I for example complain far and wide about the gen-modificated species and the inabilitiy to either give them unique names automatically or the possibility to rename them. But I don't do that everywhere but in the respective threads.
If you are posting your bug reports in the bug forum and your suggestions in the suggestions forum, you may have not that many viewers within the general usership of the forum, but you make sure that the devs take notice of them.

The real life equivalent would be complaining about the knife being blunt in a restaurant because you can opt to not get a farfalle noodle whereas you cannot play w/out sectors. What the OP has failed to grasp is that the general discontent is due to the failure to deliver after 6 months and this is the reason that discontent parties would appear all over the place.

 

[Jorgen_CAB]

The real life equivalent would be complaining about the knife being blunt in a restaurant because you can opt to not get a farfalle noodle whereas you cannot play w/out sectors. What the OP has failed to grasp is that the general discontent is due to the failure to deliver after 6 months and this is the reason that discontent parties would appear all over the place.

In my OPINION most of the complaint are conflated in regards with sectors when it clearly is an AI and design issue and not a sector issue. That is what I "try" to explain in many of my responses.

Sectors as a design feature seem quite valid to me. The Imperial government (player) are not meant to be able to directly control more than a handful of planet economies, the problem is that sectors don't have any interesting ways to interact with them, they don't have any goals so people think they are entitled to decide what they should do which I don't think was the goal with the design of sectors from the start.

I'm pretty sure sectors will get to behave even more autonomous in the the future and the difference between a protectorate, vassal or sector might blur into a grey area.

In any way I think people should try to separate what is a GENERAL AI issue versus what is a true problem with sectors. People also conflate what they like sectors to be with how Paradox intended them to be and then call that broken or a bug.

The UI leave things to be wanted, interaction with sectors could be made allot less micro (fleet planning tool would be nice). These are not bugs but design flaws.

Sectors will handle planets equally well with other AI empires so players really don't loose out on anything here, we are not suppose to control all those planets in the first place anyhow. Any work on the AI will immediately benefit not just the players sector AI but AI in general, so any work in this regard is great.

They have said they are redoing slave and purge mechanic so the AI and sector can handle them, so no point raising that anymore. We know this is not optimal currently, it does work to some extent but every player want it to work differently and they don't want to give too good tools to the players because that will make player sectors too effective versus standard AI, thus making the game too easy.

In my OPINION it is way more likely Paradox will remove player direct control from planet using planetary edicts/policies rather than allowing player more direct control over sectors. The game should in my opinion be more about macro decisions rather than micro decisions in regards to economy and politics. This would at least make the game a bit more challenging, but it would need for internal political mechanics to be a thing and overall economy expanded upon.

My general feeling is that sectors are not a finished feature and lacks a bit in substance, even if they do what they are suppose to do.

 

[Risa]

Now let's look at economic / budgeting logic.

First of all, sector AI should not emphasize on something it has no control of. That includes building spaceport for anything other than solar panel / orbital farm / observatory. Fleet capacity or ship building modules are meaningless to sector as it doesn't operate fleets. That also includes building security assets like military stations or armies or synchronized defense module. Sector does no diplomacy so it has no knowledge how valuable security assets are. Those all are responsibilities of empire AI / player. They are directly accessible to empire AI / player anyway. Sector should put resources into those aspects only when either a) there's no worthier economic project left, or b) empire AI (and maybe player) pushes an order down and there are no high-priority economic projects now.

Second, redevelopment is generally wasteful of minerals. Do it only in specific situations. Those situations include:

a) Redevelop farms (but not frontier clinic / hospital or paradise dome) when planet is fully grown and there's a lot of excess food. Bonus food from happiness should be excluded from calculation of excess food, to prevent starvation caused by happiness drop.

b) Redevelop Symbol of Unity when planet is full, all pops follow national ethos, and their ethos divergence is below -10%.

c) Redevelop normal power plant on Be stone tile to Be power plant. Similarly, redevelop tier 1 or tier 2 research lab on xeno pet tile to xeno zoo.

d) Redevelop non-unique building into power plants when sector energy income is negative, sector energy stockpile is running low, and there is no [suitable*] empty tile in all planets of the sector to construct power plants. [suitable*] here means the tile has less than +2 food / mineral / research, as well as a pop on the same planet working on non-food non-energy tile (so that it can be reassigned to new power plant once it completes).

e) When sector mineral stockpile is more than half full and no better economic option available on planet, redevelopment is fine. Mineral efficiency is not of concern in this case.


Now let's see general priorities of sector AI.

1) Ensuring each planet has at least +1 food until it's full grown shall be sector AI's top priority. When there are multiple ways to boost food production, choose one based on mineral efficiency.

2) Ensuring population growth is never stopped by removable tile blockers shall be the second top priority. This implies sector should strive to always have 100 energy ready to remove tile blocker.

3) Ensuring sector not running low on energy shall be the third top priority. Running low means has negative energy income and a stockpile lasts less than 30 months (enough time to complete whatever building on construction and then a power plant). With "Subsidize sector" function, sector running out of energy isn't disastrous end of day, so securing energy supply isn't topmost priority.

After above three, it's usually about mineral efficiency.

Tier 1 buildings for a pop on empty tile has mineral efficiency (output/mineral) of 1/30 (farm and mine) to 1/20 (energy and research). Tier 0s' are 1/30 (farm and mine) to 1/15 (energy). Tier 1 special buildings (energy hub / mineral processing plant / frontier clinic) have lower efficiency. Though even if we completely disregard their special effects, their efficiency is still at 1/75. On the other hand, upgrading building to tier 2 is 1/90, to tier 3 is 1/120, tier 4 1/150. So obviously, unless top priorities interfere, constructing new building (including three special buildings) for pop on empty tile always precedes upgrading building.

Is upgrading colony shelter when planet reaches 5 pops an exception to that rule? Upgrading colony shelter to planet administration cost 350 minerals, while provides 4 extra output and up to 4 adjacency bonus. So that's average 6 output / 350 minerals, or efficiency 1/58.3. The total efficiency of building tier 0 then upgrading to tier 1 is 1/45 (farm and mine) to 1/30 (energy). So no exception here.

What about upgrading planet administration to planet capital? It costs 500 minerals and provides 2 extra output and usually 4 extra adjacency bonus. So its efficiency is about 1/83.3. Given it also confers other benefits, its priority should be a little higher than upgrading building to tier 2.

When there's a pop working on an empty tile, barring top priorities, how to decide to construct what building on which tile, then? It should aim to maximize output of this pop. So naturally it should respect tile deposit, and favors tile with higher value ones. In cases both tier 0 and tier 1 buildings are available, tile with tier 0 building should calculate its effective deposit value with -1.5. (-1 due to 1 less output than tier 1, -0.5 due to efficiency loss in future upgrade.) Useful special buildings have higher priority than normal ones. If all other are equal, mineral > energy >= research > food.

Space mining / research station of +2 has efficiency of 1/45. Of +3, 1/30. Those are most common instances. So their (as well as constructor ship itself's) general priority is lower than new building and higher than upgrading building.

Spaceport modules themselves have good mineral efficiency comparable to tier 1 buildings, so they have similar priorities when there is empty module slot(s). However, opening slot itself, especially opening the first slot (ie. building spaceport lv 1), is quite expensive, which pushes its total efficiency to just above tier 2 buildings. In case of first slot, it's even lower than that.

Assembling robots is hard to measure by mineral efficiency. It is more energy dependent. Two things are of certain, though. One is that when there is already a pop on an empty tile, building a robot is not an efficient action. The other is that total mineral efficiency of a mine and a robot is lower than a space mining station of +2. So effectively, priority of assembling robots is lower than both tier 1 buildings and space mining / research stations.

Colony ship costs about the same as lv 1 spaceport, comes with no short term benefit, but quite a lot short term burdens (maintenance cost for ship and colony establishment phase, tech cost penalty for new world). So its priority should be at least lower than spaceport lv 1, that is, lower than tier 2 buildings. Perhaps even lower than tier 3.

As for priorities of orders from empire AI, I think it's safe to draw it below mining / research stations and above spaceport upgrade and tier 2 buildings.


How does the sector AI use all these priorities, then? Once per every months, it should check every idling planet / spaceport / constructor ship, each produces a highest priority project that affordable from possible ones to form a list, then process the list from high to low. Next month, get a new list.

 

[Reaperdamo]

So many times I've just wanted to help a planet out by clearing all its tile blockers myself using the empire's coffers but- nope- it's in a sector.

My understanding of sectors was that they were a clever idea to reduce the need for micromanagement. I actually think they serve that purpose rather well. But there is no reason to totally lock out those of us who on occasion have a need to micromanage this or that. Sector AI should help us, not replace us. By this I mean we should be able to directly control a planet whether it is in a sector or not; the sector AI would continue to improve and manage the planet while our focus is elsewhere.

+1

 

[Druesling]

Now let's look at economic / budgeting logic.

tl;wq

Wow, this isn't a very thorough look at all of the buildings stuff, the economic logic I read here will even change the way I build on new planets manually. Thanks for that!

 

[Agamemnic]

This is AI not sector. Nonetheless, you can count on the fact that sector AI is very likely doing the same thing.

I know AI coding is a nightmare but please Paradox fix this nonsense

 

[emperor_kk]

View attachment 219837
This is AI not sector. Nonetheless, you can count on the fact that sector AI is very likely doing the same thing.

I know AI coding is a nightmare but please Paradox fix this nonsense

No it's not @Risa just did all the hard work.
What Risa did was the difficult task needed, from now on it's pretty straightforward.

Now they should get the interns to follow Risa's instructions and we will have a working and efficient sector.

Thats about 10 mandays for an inecperienced user to write it, do basic testing, based on my own experirnse with clausewitch so far. So how about it PDS, half the nagging will stop if you just follow Risa's instructions.

I may not agree with some of his decisions but if you follow his instructions I'm sure most of the people will shut up.

 

[Jorgen_CAB]

IMO when people say something like "have u ever try to give sector 5k minerals?", they fundamentally wrong.
Lets try to understand why with example of such game as SC2: when u hign on minerals(like >500), you are playing bad. The reason is that you wasting potential buildings or army. So I have a question for all of you: for which purpose sectors want to save thousands of minerals, if they cant build wa
rships and basic mines costs ~70 minerals?
This is economically wrong even for AI empires: if u saving minerals not for expensive important building\upgrade\ship, you're performing bad. Keep stockpiles as empty as possible.

This is how the AI work... the AI earmark minerals in different areas for different things and that is why it needs a good storage of minerals to start building. The amount it needs depend on many factors such as storage size, technology and size of the sectors as well I guess.

All AI work this way and this is one reason why it take so long for the AI to do stuff, it also is rather bad at building up its initial economy in an efficient way.

A player obviously use his minerals as fast as he can, a human player can keep those earmarked stores in his head and imagine how future mined minerals can be spent. You might eventually want to store both energy and minerals so you have a reserve during wars, but other than that you don't want to have any huge mineral stores. The way the game works you always want to use minerals as fast as possible.

 

[emperor_kk]

This is how the AI work... the AI earmark minerals in different areas for different things and that is why it needs a good storage of minerals to start building. The amount it needs depend on many factors such as storage size, technology and size of the sectors as well I guess.

All AI work this way and this is one reason why it take so long for the AI to do stuff, it also is rather bad at building up its initial economy in an efficient way.

A player obviously use his minerals as fast as he can, a human player can keep those earmarked stores in his head and imagine how future mined minerals can be spent. You might eventually want to store both energy and minerals so you have a reserve during wars, but other than that you don't want to have any huge mineral stores. The way the game works you always want to use minerals as fast as possible.

No this is how poorly written, badly designed, non optimized, half tested things work. This is a game AI, so at best, a fuzzy logic path-finding equivalent. You only need to figure out the set of rules by which it decides and it will do its thing as long as you keep it out of a loop or a non recovering endwall scenario. The only reason it stores before using any of it is because there is no escape path if it rapidly goes into negative and it will just stay there. A sloppy solution for not being able to handle resource starvation.

 

[Jorgen_CAB]

No this is how poorly written, badly designed, non optimized, half tested things work. This is a game AI, so at best, a fuzzy logic path-finding equivalent. You only need to figure out the set of rules by which it decides and it will do its thing as long as you keep it out of a loop or a non recovering endwall scenario. The only reason it stores before using any of it is because there is no escape path if it rapidly goes into negative and it will just stay there. A sloppy solution for not being able to handle resource starvation.

I suppose you did not understand I just explained how it worked, not if I thought it was any good way to deal with it. When I referenced ALL AI I was talking about Stellaris general AI not AI in other games.

I'm a programmer myself and I actually programmed AI a long time ago, I would not have done it this way... but I had a bit more resources to work with when I did it... so there always are restrictions you need to deal with. CPU are only that great.

I would use a dynamic and weighted priority system for spending resources and put in breaks for how deficit spending can occur and how it needs to be dealt with. No need for the AI to store minerals when it can do that virtually without actually having the resources.

 

[emperor_kk]

I suppose you did not understand I just explained how it worked, not if I thought it was any good way to deal with it. When I referenced ALL AI I was talking about Stellaris general AI not AI in other games.

I'm a programmer myself and I actually programmed AI a long time ago, I would not have done it this way... but I had a bit more resources to work with when I did it... so there always are restrictions you need to deal with. CPU are only that great.

I would use a dynamic and weighted priority system for spending resources and put in breaks for how deficit spending can occur and how it needs to be dealt with. No need for the AI to store minerals when it can do that virtually without actually having the resources.

You were too polite, I chose not to be. A dynamic weighted priority system based on consumption allotments would be awesome but that would never be possible here. Both bvecause the scripting engine has limits to what you can do in a cost effective way and because judging by what i've seen so far in the scripts the level is at best mediocre or extremely high done in a few hours worth of time just so it can ship out. Whatever the case, that would be too complicated to be implemented by this studio. I do not have such high hopes.

Also I tend to disfavour very dynamic systems when i have runtime restrictions. For example I would train a neural network for optimizing a purchase/stock/sales alogorythm in a non standard heurestic calculation for stock control. But I would only ship out the finished algorythm with pre-defined optimization subroutines for specific parameters change so as to lose the heavy duty calculation that would make it purely dynamic and only take advantage of the end result, so as to be able to produce result with low-end office pc.

 

[AlphaAsh]

"...A dynamic weighted priority system based on consumption allotments would be awesome but that would never be possible here..."

It is. There's lots of other run-time queries that can be made that aren't CPU hungry.

I'm going to be somewhat blunt (again) about this and this time I am going to apologise in advance to the AI programmer(s) working on Stellaris, but they do have the tools to do better. I just don't think they're being given enough time or resources to do better.

Why do I say this? I'm pouring over AI weighting soft-code for planet development and getting some startlingly good results in AlphaMod with the latest.

Also, take a look at said soft-code for vanilla buildings in some of the latest patches. Someone in the team is trying to improve the relevant AI behaviour with it. I still think it's very lacking though, especially if the back-end continues to be as piss-poor as it is.

And why do I say that? One example: decisions about whether to build happiness buildings are being fronted to soft-code. That right there is an indication that the back-end AI programmer(s) isn't getting the time to do what he/she/they need to do.

EDIT - However, there's a core issue to the AI right now, and that is the very basic decision as to when to allocate resources to different areas. The Ai currently behaves as if developing a planetary infrastructure is low priority when in fact it is the highest priority. If you don't have planets growing pops to produce energy and minerals, then your economy is going to death-spiral. Without that infrastructure, pouring out ships for fightses is going to get harder and harder.

 

[Robert_Sheckley]

They cant teach it to build "habitat clinic" on food resource when available for 7 month I would fix this and many others glitches in 3 days. Is paradox hiring some kind of cheap Turkish immigrants to work on the AI?

 

[Tim_Ward]

View attachment 219837
This is AI not sector. Nonetheless, you can count on the fact that sector AI is very likely doing the same thing.

I know AI coding is a nightmare but please Paradox fix this nonsense

Ah, the classic Stellaris AI 'about 20 more food than the planet needs' there.

 

[Tim_Ward]

They cant teach it to build "habitat clinic" on food resource when available for 7 month I would fix this and many others glitches in 3 days. Is paradox hiring some kind of cheap Turkish immigrants to work on the AI?

Very much not cool.

 

[Elfin]

Personally I don't have much problems with sectors (although that are mostly "problems with AI"). But there is one thing I really don't understand - why do we have sector limit? Even in my largest empires I had only 3-4 sectors, some of them containing up to 100 habitable planets. This is weird. Maybe you should add sort of "planet limit for sector"?

 

[Druesling]

Personally I don't have much problems with sectors (although that are mostly "problems with AI"). But there is one thing I really don't understand - why do we have sector limit? Even in my largest empires I had only 3-4 sectors, some of them containing up to 100 habitable planets. This is weird. Maybe you should add sort of "planet limit for sector"?

Yes. That's curious. Maybe there once was an exploit where you put every colony in its own sector, but I don't think that's working anymore in 1.3, if it ever did.

 

[Jorgen_CAB]

"...A dynamic weighted priority system based on consumption allotments would be awesome but that would never be possible here..."

It is. There's lots of other run-time queries that can be made that aren't CPU hungry.

I'm going to be somewhat blunt (again) about this and this time I am going to apologise in advance to the AI programmer(s) working on Stellaris, but they do have the tools to do better. I just don't think they're being given enough time or resources to do better.

Why do I say this? I'm pouring over AI weighting soft-code for planet development and getting some startlingly good results in AlphaMod with the latest.

Also, take a look at said soft-code for vanilla buildings in some of the latest patches. Someone in the team is trying to improve the relevant AI behaviour with it. I still think it's very lacking though, especially if the back-end continues to be as piss-poor as it is.

And why do I say that? One example: decisions about whether to build happiness buildings are being fronted to soft-code. That right there is an indication that the back-end AI programmer(s) isn't getting the time to do what he/she/they need to do.

EDIT - However, there's a core issue to the AI right now, and that is the very basic decision as to when to allocate resources to different areas. The Ai currently behaves as if developing a planetary infrastructure is low priority when in fact it is the highest priority. If you don't have planets growing pops to produce energy and minerals, then your economy is going to death-spiral. Without that infrastructure, pouring out ships for fightses is going to get harder and harder.

I agree... the underlying AI is not that advanced and the scripted AI is rather slow so making that any heavier are not really much an option.

What I was getting at was creating a much more dynamic system in the source code, at least for the basic stuff such as food and happiness and so forth... problem obviously is as always... different player want different things at different times. But I believe the AI in the source code could do better.

I have experimented quite allot with the scripted AI and I find it a rather blunt instrument and it is a bit to 1 or 0 oriented. There are not enough input that can impact the AI decision or the script code become miles long.

 

[Jorgen_CAB]

Personally I don't have much problems with sectors (although that are mostly "problems with AI"). But there is one thing I really don't understand - why do we have sector limit? Even in my largest empires I had only 3-4 sectors, some of them containing up to 100 habitable planets. This is weird. Maybe you should add sort of "planet limit for sector"?

Yes... this is weird to me too... with experimenting I find that sectors work best with only a handful of systems in them, perhaps 4-7 but no more than 10. Problem is of course governors when you have a really big empire. It is also weird that sectors can manage that many more planets than the imperial core. Perhaps twice as many would be acceptable but not hundreds of planets or systems in one sector.

I think they will redo how sectors work in a major way at some point.

 

[Risa]

EDIT - However, there's a core issue to the AI right now, and that is the very basic decision as to when to allocate resources to different areas. The Ai currently behaves as if developing a planetary infrastructure is low priority when in fact it is the highest priority. If you don't have planets growing pops to produce energy and minerals, then your economy is going to death-spiral. Without that infrastructure, pouring out ships for fightses is going to get harder and harder.

To be fair, decision on military or infrastructure is always one of weakest link in AI of 4x games. I've never seen a game whose AI arrives close on balance point. Hell, even human players need a lot of trial and error to discover the most effective way, and that is most probably just best against AI in that game only. Copying it into multiplayer is more often than not an inferior strategy.

That's why my suggestion on budget logic aims to sector AI only. Empire AI has to face that difficult problem. Sector AI, on the other hand, has not and should not.