Hello everyone and welcome to another Stellaris development diary. Today's dev diary will be focusing on the road ahead after Cherryh and Apocalypse, and our long-term priorities going forward.

Cherryh Post-Release Support
As mention in last week's dev diary, the immediate priority for the team is post-release support for Cherryh and Apocalypse, fixing bugs, addressing balance/feature feedback, and working on quality of life and performance improvements. We are maintaining a running 2.0.2 beta patch which we will continue to update every few days or so until we are happy with the state of the game.

The Post-Apocalypse
Apocalypse and Cherryh were an expansion/patch focused almost exclusively on war, and with it out, we are now going to be moving on to other, non-war related priorities for future updates, expansions and story packs. To give you an idea of what's coming, we're going to revisit the list of long-term goals for Stellaris I made and updated for Dev Diary #50 and Dev Diary #69. This time, we're going to organize the goals into the ones we feel have been delivered on, old goals that were added to the list before 2.0, and new goals that we have set for ourselves after 2.0 (there is no prioritization difference between goals based on when they were added or whether they are considered old or new for this particular list).

As before, the list is NOT in order of priority, and something being considered completed NOT mean we aren't going to continue to improve on it in future updates, just that we consider it to be at a satisfactory level.

As before, THIS IS NOT AN EXHAUSTIVE OR FINAL LIST, NOTHING NOT ALREADY COMPLETED IS CERTAIN TO HAPPEN AND THERE ARE NO ETAS

Completed Goals

• Ship appearance that differs for each empire, so no two empires' ships look exactly the same.
• More potential for empire customization, ability to build competitive 'tall' empires.
• Global food that can be shared between planets.
• Ability to construct space habitats and ringworlds.
• Factions that are proper interest groups with specific likes and dislikes and the potential to be a benefit to an empire instead of just being rebels.
• Ability to set rights and obligations for particular species in your empire.
• Buildable Dreadnoughts and Titans.
• Deeper mechanics and unique portraits for synthetics.
• Reworking the endgame crises to be more balanced against each other and the size/state of the galaxy.
• Reworks to war to address the 'doomstacks' issue and make the strategic and tactical layers of warfare more interesting and less micro-intensive.
• Superweapons and planet killers.

Old Goals

• A 'galactic community' with interstellar politics and a 'space UN'.
• Deeper Federations that start out as loose alliances and can eventually be turned into single states through diplomatic manuevering.
• More story events and reactive narratives that give a sense of an unfolding story as you play.
• More interesting mechanics for pre-FTL civilizations.
• 'Living systems', making empire systems feel more alive and lived in

New Goals

• Less micromanagement and more focus on interesting choices in regards to planets, the ability to grow planets beyond current fixed size.
• Empire trade mechanics and trade agreements.
• A galactic market where resources and strategic resources can be imported and exported.
• Espionage and sabotage mechanics.
• Improved galaxy/hyperlane generation with better placed systems and dangers.
• More anomalies and unique systems to explore.

That's all for today! Over the next few weeks, dev diaries will continue to focus on post-release support. Feature dev diaries will resume when we have new features to talk about. Finishing off this dev diary is a screenshot of how we're reworking difficulty modes in the next update to the rolling 2.0.2 beta:

 

[IlyaZ]

I don't think you realize how resource intense a decent neural net that can actually stand up to the current game's "AI" would still be, aside from needing a supercomputer to make it (and it would take a long while for something decent alongside being likely out of budget range), its also not that feasible for the player in any account. Neural nets especially have an issue of providing inconsistent behavior for things like games if they're worth anything, that's what makes highly evolved ones so challenging but for a video game that's actually a terrible trait because games need consistent behavior in "AI" and making something that's always unpredictable basically equates to either doing something stupid or doing something bullshit. (the former we already yell at the current "AI" for and the latter fundamentally destroys the game as the player feels they're being screwed a fair challenge) You would also need to run multiple non-player empires through the neural net which depending on a lot of things can become expensive, and its not even guaranteed what is to happen in any case. Using true AI techniques to create game "AI" isn't done because there are too many problems and true AI doesn't remotely share the same purpose nor will it contain the same goals as game "AI" and I really wish people would stop trying to equate them.

It probably is very resource intense, but in principle it only has to be done once. The problem is that the DLCs and patches change a lot of things regularly. To mimic a human player (fog of war, no cheats and so on) you will have to deal with an adaptive opponent that learns from games eventually. For basic AI, it's probably much easier to do hard coded logic and you get better results, but if players want something more than standard current gen AI I don't see any other way forward than artificial neural nets as has been successfully implemented through the use of supercomputers in other RTS. The anti-neural net and evolutionary algorithm sentiment is understandable for most commercial applications, but in order to get beyond that the adaptation of new technology is absolutely necessary.

 

[Bearjuden]

since when you find a Precursor home system it does just seem to plonk a new system-plus-hyperlane somewhere.

It does? I just assumed it existed when the game begins but it hid them somehow. Then the map is constant. Seems much simpler to code, too, so I figure that's probably what Paradox did.

 

[Tyloo]

• 'Living systems', making empire systems feel more alive and lived in
• Less micromanagement and more focus on interesting choices in regards to planets, the ability to grow planets beyond current fixed size.

In regards to these two points, I think it would be really cool if we could set up small micro colonies on "non-habitable" planets in our colonized systems. Think of it like setting up a colony on Mars or something. While we can't support a totally colonized planet because we can't terraform it, we can bring and transport enough supplies to develop a small specialized outpost on these planets. Only able to be constructed in systems with a proper colony, they would add 1 - 4 tiles and perhaps wouldn't need a planetary administration.
That way, there would be a reason to add smaller ships (that are only aesthetic) flying between places in the systems in order to transport resources or mine the asteroid belt etc. Same thing could be done for transporting resources out of existing mining stations etc. No changes to gameplay here, just something to give systems some life.

 

[KubiG37]

It's a 4X. Symmetrical game mechanics give the player a huge advantage.

Then why AIs in games like Europa IV or Victoria don't get any flatout cheating bonuses too? (except some to simply reduce cpu workload) Aren't they Paradox games too? (and much more complex to be fair)
I really hate it when AIs outright cheat, and by forcing them to not cheat and simply play fair (without abusing their own faults from my side) and then being called "beginner, newbie, or ensign" is just wrong. I just dislike games where single planet/base AIs can afford armies rivaling that of your 4 times larger nation. (like in all Total War games for example)

 

[Lilhelpy]

Then why AIs in games like Europa IV or Victoria don't get any flatout cheating bonuses too? (except some to simply reduce cpu workload) Aren't they Paradox games too?

They don't have tile system.

 

[Bearjuden]

Then why AIs in games like Europa IV or Victoria don't get any flatout cheating bonuses too? (except some to simply reduce cpu workload) Aren't they Paradox games too? (and much more complex to be fair)

To be even more fair, EU4 is a straight up 4X game, so it's not much of a counter-argument. It meets all the criteria. (Victoria II is not, though, although really the problem is not limited to only 4X games; any system with defined and static rules - ie nearly any videogame - gives the player an advantage, since you can think and experiment, while it...just does what it was told).

They don't have tile system.

This doesn't matter. The point isn't that they use a specific system. The point is that they use the same system as everyone else.

 

[Lilhelpy]

This doesn't matter. The point isn't that they use a specific system. The point is that they use the same system as everyone else.

Well, it does. The AI is currently bad at working the tiles, and it is tiles that make economy, and it is economy that pumps out battleships.
Those 'cheats' basically just help AI to handle tiles. Developers just need an accountant to do the scripting and AI tile management, not programmer, then, maybe, it won't need cheats

 

[Bearjuden]

Those 'cheats' basically just help AI to handle tiles. Developers just need an accountant to do the scripting and AI tile management, not programmer, then, maybe, it won't need cheats

It's not that simple. An accountant can figure out what is the best order to build buildings in to maximize one resource easily, two resources pretty reliably. But there are many resources to keep track of - besides the obvious minerals and energy, you have food, unity, fleet power, leaders. Every time you build a destroyer, you're putting off a mine by another few months; every time you hire a leader, you aren't clearing that tile blocker. Add in to that military tactics - the AI has to know how to spend those resources on defense as well, which is not even easily measurable - and this becomes a job far in excess of what a mere accountant can do.

 

[Lilhelpy]

It's not that simple.

True. It is a very complex system of priorities and decisions. To make AI good with all that stuff requiers a lot of work. Maybe years of work to tune every aspect of AI economy. If I will write a line for every economical decision I make in game, including every check using only yes/no arguement, I will go crazy withing an hour probably

Yes, sometimes it makes me sad when I look at AI planets. But again, until we have an AI building economy properly, it's not bad when it cheats. The team of mad proggrammer-accountants at Paradox will fix it one day. And we can enjoy the game right now.

 

[FleetingRain]

Did you guys just say EUIV has no AI cheats lmao

 

[tangled axile]

Then why AIs in games like Europa IV or Victoria don't get any flatout cheating bonuses too? (except some to simply reduce cpu workload) Aren't they Paradox games too? (and much more complex to be fair)
I really hate it when AIs outright cheat, and by forcing them to not cheat and simply play fair (without abusing their own faults from my side) and then being called "beginner, newbie, or ensign" is just wrong. I just dislike games where single planet/base AIs can afford armies rivaling that of your 4 times larger nation. (like in all Total War games for example)

To be fair, it doesn't specifically call you an ensign. You could take it that way, sure, but you could just as easily consider it to be describing the AI. That's more consistent with the "scaling" setting, anyway.

Sure it's semantic, but they DIDN'T call it "beginner" or "noob" or whatever; the game actually isn't insulting you directly like you suggest. See Wolfenstein for a case of that.

Also, isn't this the first time pds has used a tile system? If so, of course they're not as good at programming the AI to use it compared to the systems of their other games...

 

[Primarch Victus]

"Espionage and sabotage mechanics."



YEEEEESSSSS!!!

 

[MPixels]

Then why AIs in games like Europa IV or Victoria don't get any flatout cheating bonuses too? (except some to simply reduce cpu workload) Aren't they Paradox games too? (and much more complex to be fair)

I said 4X. Victoria and EU are grand strategy games. Crucially, they do not have symmetrical starts, typical of 4X gameplay (although I recall Wiz mentioning years ago that the AI has no naval attrition and some colonisation buffs to make it competitive). I hope I've clarified what you got confused by misrepresenting.

 

[Bearjuden]

I said 4X. Victoria and EU are grand strategy games. Crucially, they do not have symmetrical starts, typical of 4X gameplay (although I recall Wiz mentioning years ago that the AI has no naval attrition and some colonisation buffs to make it competitive). I hope I've clarified what you got confused by misrepresenting.

Said it many times before, and I'll say it again, EU is a 4x game that won't admit it to itself. Having symmetrical starts is a convention of 4x gaming, not a rule. The only thing that matters is those 4 x's and EU has them all.

That being said, I definitely like it when there is the option for no AI bonuses. Sometimes you want a challenge, sometimes you want a sandbox. Just let the difficulty be set to whatever the player wants. If you want to restrict achievements to only "neutral" and higher, that's fine, just so long as there's the choice.

 

[MPixels]

Said it many times before, and I'll say it again, EU is a 4x game that won't admit it to itself. Having symmetrical starts is a convention of 4x gaming, not a rule. The only thing that matters is those 4 x's and EU has them all.

I understand this logic but the conventions are what make it a meaningful genre term. Portal is first-person and you shoot a gun. Is it an FPS? Furthermore, those 4 Xes apply to RTS games like Age of Empires but you'd look mad if you called them 4X strategy games.

 

[Bearjuden]

I understand this logic but the conventions are what make it a meaningful genre term. Portal is first-person and you shoot a gun. Is it an FPS? Furthermore, those 4 Xes apply to RTS games like Age of Empires but you'd look mad if you called them 4X strategy games.

Mad, maybe, but not wrong. It's helpful for game design to know whether it is or isn't. 4x games have always suffered from a positive feedback problem - winning a fight gives you a reward, and the reward makes it easier to win future fights. At some point, you've won the hardest fight you'll have, and the rest of the game is trivial. If that sounds disturbingly like EU4, Age of Empires, or Civ, there's a reason for that, and there's a reason both EU and Civ give the AI huge bonuses on the settings you're most likely to play (AoE doesn't give any flat modifiers to my knowledge, but their scenarios at least are always built such that you are at a hefty material disadvantage, accomplishing the same thing).

Portal's a bit of a weird example, I will admit, because it's both first person but also a puzzle game; I would argue that it deserves a separate category for things like it, Antichamber, Quantum Conundrum, etc (Puzzle Gimmick games?). They all fall into a similar pattern - you have a special tool (the gimmick), you learn how to use it (the puzzle), the tool gets more complicated, you solve more complicated puzzles with it, and this keeps escalating until the game ends and you know exactly what the tool's limitations are and how it can be best applied.

 

[MPixels]

Mad, maybe, but not wrong...

But you see my point about strict narrow definitions for genres? If 4X or FPS applies for a great variety of games, which have their own labels (i.e. RTS covers most MOBAs but we reserve the latter term for that genre) it makes things a lot clearer?

 

[Spartan322]

It probably is very resource intense, but in principle it only has to be done once. The problem is that the DLCs and patches change a lot of things regularly. To mimic a human player (fog of war, no cheats and so on) you will have to deal with an adaptive opponent that learns from games eventually. For basic AI, it's probably much easier to do hard coded logic and you get better results, but if players want something more than standard current gen AI I don't see any other way forward than artificial neural nets as has been successfully implemented through the use of supercomputers in other RTS. The anti-neural net and evolutionary algorithm sentiment is understandable for most commercial applications, but in order to get beyond that the adaptation of new technology is absolutely necessary.

I have my doubts on some of the game "AI" being true neural nets, but regardless, RTS games are relatively much easier to get neural nets to train, problem is would they actually do a better for the gameplay, and I'd say without a lot of money, time, and a very rigid set of narrow rules to follow, nets will generally suck. Now get past that it becomes nearly impossible to tune the net to fit a proper balance, you might get something that is exceedingly dumb even after a lot of iterations or you could get something that is too smart (given you put a lot of time, effort, and had a rigid model) and wouldn't feel engaging to play against. That means it works best for a linear perspective on tactics in which it will make predictions and qualify whether it can win or lose. As far as would it be able to build decently that would require at least another net if not a few. And then you gotta separate that to each empire. Looking at this issue not only do you have to train the NN on multiple subsets of mechanics and have it act independently both of those mechanics and of other empires, but you gotta spend a fortune, a lot of time, and even then the results aren't even ensured when its finished, getting NNs to do what you want is very hit or miss and given we see it be hard enough to get it between smart and dumb enough to be engaging to play that means its a huge risk, and I'd think with actually very vague and shallow rewards. And that doesn't really mean it will do everything better anyway, so it gets better at simulating a player fighting a war, that's cool but does it mean it can reasonably manage its energy and minerals without cheating? I don't know but trying to even figure that out is likely worth less then just scripting behaviors to do it properly, and its probably more efficient in manner of money, time, and power. (computer learning algorithms are not an end all be all for adaptable AI [they don't actually often work for game "AI" because they're not really suppose to be AI] especially since it will need to be fixed with every tiny change and then a training process could produce an even dumber version and not a smarter one)

 

[Bearjuden]

But you see my point about strict narrow definitions for genres? If 4X or FPS applies for a great variety of games, which have their own labels (i.e. RTS covers most MOBAs but we reserve the latter term for that genre) it makes things a lot clearer?

...possibly? What I would definitely say is that it's not only important to realize how your game is similar to others in the same genre but also different, and account for all the impacts of both, which is possibly a very drawn out way to say what I think you're saying.

 

[Hype]

May I make a suggestion?

Back in the changelog of the 2.0 update (Dev Diary #105):



While I see why this was implemented, it can lead to some awkward situations, such as in the Ransomeers event chain, when the ransomeer space station spawns in a system right next to your borders that you don't have any intel on, meaning you have to pull your science ship from halfway across the galaxy all the way back to your borders just to enter this system, possibly risking your ship and your scientist in the process.

Perhaps events like these could be adjusted in a way that makes sense? Like the captives finding a way to send a scan of the system to you, letting you send your fleets in without having to send in a science ship first.

How do you have all your science ships halfway across the galaxy and still have systems right next to your borders with no intel? Early game you need to scout those systems and after early game sensors will give you intel on those systems so you won't need a science ship.