Fractal Softworks Forum

Please login or register.

Login with username, password and session length
Advanced search  

News:

Starsector 0.97a is out! (02/02/24); New blog post: Simulator Enhancements (03/13/24)

Pages: [1] 2 3

Author Topic: Procedural Generation needs improvement  (Read 8749 times)

Sendrien

  • Captain
  • ****
  • Posts: 265
    • View Profile
Procedural Generation needs improvement
« on: January 04, 2019, 01:51:38 PM »

In general, I have only praise for how Starsector has progressed through the versions. However, the recent additions in 0.9 have brought to light something that I didn't really notice earlier, but now is becoming painfully obvious.

Procedurally generated solar systems are somehow a) emptier; b) duller and/or c) less interesting than the hand-crafted ones in the Core Worlds.

For instance, the vast majority of habitable planets are found in simplistic systems which will include: 1 star, 1-3 stable points, 1-3 planets, 1 astronomical feature (ring system or magnetic field) and the accompanying jump points. In many cases, the system is completely empty! Just a star with no planets! I realize that this may be representative of the emptiness of space, but perhaps we should think of the sector map as a map of *interesting* systems? In terms of player experience, would it not be much more beneficial if each system were worth exploring?

Contrast this to the complex and extremely interesting features of practically every system in the Core Worlds! Some planets have multiple moons, asteroid belts, and a whole bunch of other quirky features that makes each location memorable.

It really kills the "Eureka moment" of finding a Terran planet when you realize that it is literally the only planet orbiting a completely empty system.

Is there some kind of memory limitation or limit on the number of planetary bodies that somehow requires the procedurally generated systems to be so simplistic? Otherwise, I would like to suggest that the parameters be tweaked so that most systems are more complex and contain between 3 and 18 planetary bodies, including moons.

Anecdotally, one of the most awesome moments of discovery I've encountered in Starsector was my first encounter with a Pulsar. Being pushed away by the Pulsar was wicked. Seeing the pulsar's beam of light be dissected by an orbiting planet...just awesome. You know what the big letdown of that system was? It was completely empty save for that one planet.
« Last Edit: January 04, 2019, 02:41:51 PM by Sendrien »
Logged

Alex

  • Administrator
  • Admiral
  • *****
  • Posts: 23987
    • View Profile
Re: Procedural Generation needs improvement
« Reply #1 on: January 04, 2019, 02:51:23 PM »

For instance, the vast majority of habitable planets are found in simplistic systems which will include: 1 star, 1-3 stable points, 1-3 planets, 1 astronomical feature (ring system or magnetic field) and the accompanying jump points. In many cases, the system is completely empty! Just a star with no planets! I realize that this may be representative of the emptiness of space, but perhaps we should think of the sector map as a map of *interesting* systems? In terms of player experience, would it not be much more beneficial if each system were worth exploring?

With the caveat that some of this is personal preference, I think it's actually quite important for a decent number of systems to be empty or relatively uninteresting. It establishes a baseline for something to be "interesting" in comparison to. If everything is "worth exploring", it'll take away a lot of the fun of exploration and make it feel more like a chore. Contrast that with finding a system that's actually worth exploring, after seeing a number that aren't.

There are also lots of ways the game steers the player towards the more interesting systems, from something relatively subtle (special names) to missions, warning beacons, etc, so the player is given tools to explore more (or less) wisely, which the less interesting systems make possible.

That's not to say procgen is perfect and might not benefit from a tweak here and there. I just disagree with the premise that everything needs to be "interesting"; if everything is, then nothing is, you know?
« Last Edit: January 04, 2019, 02:59:48 PM by Alex »
Logged

From a Faster Time

  • Commander
  • ***
  • Posts: 146
    • View Profile
Re: Procedural Generation needs improvement
« Reply #2 on: January 04, 2019, 03:28:33 PM »

That's not to say procgen is perfect and might not benefit from a tweak here and there. I just disagree with the premise that everything needs to be "interesting"; if everything is, then nothing is, you know?
Agreed, the spice, the specials stuff should be special. As long as it's not super rare to the point where there is only 5 interesting systems in the whole game.
Logged

Sendrien

  • Captain
  • ****
  • Posts: 265
    • View Profile
Re: Procedural Generation needs improvement
« Reply #3 on: January 04, 2019, 03:39:55 PM »


With the caveat that some of this is personal preference, I think it's actually quite important for a decent number of systems to be empty or relatively uninteresting. It establishes a baseline for something to be "interesting" in comparison to. If everything is "worth exploring", it'll take away a lot of the fun of exploration and make it feel more like a chore. Contrast that with finding a system that's actually worth exploring, after seeing a number that aren't.

I can see your point about interesting becoming normal. However, really special systems which contain habitable planets or other special features like Pulsars, perhaps could carry some kind of "Special" tag, so that the procedural generator gives them more substance.

Don't forget that the estimate of *average* number of planets per star in the Milky Way is around 8, with a low and high range of 2 and 80. (Source: https://www.universetoday.com/30296/how-many-planets-are-in-the-galaxy/)

I consider myself very lucky if I find 1 procedurally generated system in a given save with at least 6 planetary bodies (including moons) in Starsector.

What about giving systems denoted with the "Special" tag a 40% chance to generate 6 planets or more?
Logged

Alex

  • Administrator
  • Admiral
  • *****
  • Posts: 23987
    • View Profile
Re: Procedural Generation needs improvement
« Reply #4 on: January 04, 2019, 04:34:08 PM »

Hmm - just made a new game, randomly sampling a few systems, counting planets:
9, 2, 7, 1, 2, 6, 4, 1, 2, 7

Actually, I've got a thing that prints the overall stats, which I was using when fine-tuning procgen. So, for this just-now-generated Sector, we've got:
Systems: 190
Planets: 513
Habitable planets: 55

An average of 2.7 planets per system, but considering some are empty and it's a normal distribution otherwise, there'll be a fair bit of systems on the mid-to-high end.

The blue giant systems to tend to have more stuff, though, and less habitable worlds, so in general, a habitable world in a system that also has lots of other planets is less likely.

I consider myself very lucky if I find 1 procedurally generated system in a given save with at least 6 planetary bodies (including moons) in Starsector.

That seems like exceedingly bad luck! ... or, possibly, selective memory, if those systems didn't have much else of note going on :)

However, really special systems which contain habitable planets or other special features like Pulsars, perhaps could carry some kind of "Special" tag, so that the procedural generator gives them more substance.
...
What about giving systems denoted with the "Special" tag a 40% chance to generate 6 planets or more?

What's the gameplay reasoning here? I'm not saying this would necessarily be bad, but at first glance this seems like a fairly lateral change, if that makes sense - neither better nor worse, just different.
Logged

Sendrien

  • Captain
  • ****
  • Posts: 265
    • View Profile
Re: Procedural Generation needs improvement
« Reply #5 on: January 04, 2019, 06:05:27 PM »

I suppose this is a subjective judgement, but for me, one of the reasons I love Starsector is the thrill of finding something special. Finding a Terran planet from which you can build a large population center for your faction is one of those moments. You might be hunting and exploring the whole game for that Terran planet in the Sector.

Since this is one of the implicit payoffs for the amount of time and effort put into extensive exploration, I feel like it's a letdown when the Terran planet is alone in an empty system, just orbiting a star.

(It would be interesting if you did the same sampling test on the systems with Terran planets (or even habitable planets). I'd bet your numbers would be much lower, and you might find that Terran planets are often found in systems with 3 planetary bodies or fewer.
Logged

Megas

  • Admiral
  • *****
  • Posts: 12118
    • View Profile
Re: Procedural Generation needs improvement
« Reply #6 on: January 04, 2019, 08:10:55 PM »

Most of the interesting systems are the cooler stars that can have Terran or other low hazard worlds with all of the resources.

Big systems with mostly high hazard planets are... yuck!  They are only good if planets have ruins (for tech mining) or have an abandoned station to loot.
Logged

CrashToDesktop

  • Admiral
  • *****
  • Posts: 3876
  • Quartermaster
    • View Profile
Re: Procedural Generation needs improvement
« Reply #7 on: January 04, 2019, 08:21:50 PM »

Most of the interesting systems are the cooler stars that can have Terran or other low hazard worlds with all of the resources.

Big systems with mostly high hazard planets are... yuck!  They are only good if planets have ruins (for tech mining) or have an abandoned station to loot.

From a strictly min-maxing perspective.

Don't let that one perspective cloud your vision of what a visually interesting system is, either.  I find Blue Giant systems with loads of planets to be the most fun to explore - I wish more stars had larger planetary systems in them.
Logged
Quote from: Trylobot
I am officially an epoch.
Quote from: Thaago
Note: please sacrifice your goats responsibly, look up the proper pronunciation of Alex's name. We wouldn't want some other project receiving mystic power.

Deshara

  • Admiral
  • *****
  • Posts: 1578
  • Suggestion Writer
    • View Profile
Re: Procedural Generation needs improvement
« Reply #8 on: January 04, 2019, 08:27:25 PM »

my only complaint about the procgen is that there's a lot of room for extraordinarily surprising stuff hidden inside systems that look like they should be boring but the current system only does... well, as you said Alex, aside from Empty it's mostly just Normal Distribution. The game feels like it needs the occasional outlier where the player can get rewarded above-grade for exploring systems that lack the possibility of colonization.
Like, a chance for exploration content in systems that are uncolonizable to be upgraded into even better/more interesting content just for being in a system that has no value for colonies, so that exploring nebulas and systems with no planets doesn't feel like such a waste of time.
Logged
Quote from: Deshara
I cant be blamed for what I said 5 minutes ago. I was a different person back then

Alex

  • Administrator
  • Admiral
  • *****
  • Posts: 23987
    • View Profile
Re: Procedural Generation needs improvement
« Reply #9 on: January 04, 2019, 09:07:55 PM »

I suppose this is a subjective judgement, but for me, one of the reasons I love Starsector is the thrill of finding something special. Finding a Terran planet from which you can build a large population center for your faction is one of those moments. You might be hunting and exploring the whole game for that Terran planet in the Sector.

Since this is one of the implicit payoffs for the amount of time and effort put into extensive exploration, I feel like it's a letdown when the Terran planet is alone in an empty system, just orbiting a star.

I guess I'm not seeing a fundamental difference. More planets potentially make the system more useful, but, I mean, it's already got a terran planet! Beyond that, just feel-wise, I think both are interesting, and I'm not really seeing "more planets" as a win in terms of feel or reward. I think maybe it is a subjective thing, since it seems like you are.

(It would be interesting if you did the same sampling test on the systems with Terran planets (or even habitable planets). I'd bet your numbers would be much lower, and you might find that Terran planets are often found in systems with 3 planetary bodies or fewer.

I'm sure you're right, as I was saying, the larger planet-count systems tend to be blue giants, which are less likely to get habitable worlds.

(Actually, not that hard to check: maximum of 4 planets in a system with a terran in the current save slot. But it's not super representative, probably, since there's only 2 terrans total in this one.)
Logged

Sendrien

  • Captain
  • ****
  • Posts: 265
    • View Profile
Re: Procedural Generation needs improvement
« Reply #10 on: January 04, 2019, 10:08:13 PM »


I guess I'm not seeing a fundamental difference. More planets potentially make the system more useful, but, I mean, it's already got a terran planet! Beyond that, just feel-wise, I think both are interesting, and I'm not really seeing "more planets" as a win in terms of feel or reward. I think maybe it is a subjective thing, since it seems like you are.

Is the goal to have balanced utility between systems? If yes, perhaps the current approach is the way to go.

However, if we acknowledge that blue star systems are mainly going to be used for exploration, whereas the player will likely jump at the opportunity to colonize a terran planet wherever practical and possible, then from a pure gameplay perspective, those systems in which the player invests the most time and resources building up should absolutely be more visually interesting, more expansive and content-rich.

It's the same reason the Core Worlds were hand-crafted. It's the same reason in some MMOs, player hubs and boss fights are extremely well fleshed out, whereas a random filler dungeon may be a rehash of an earlier dungeon.

To take an extreme example, imagine there was only 1 Terran world in a particular save. And the procgen made that Terran system literally 1 star + Terran planet + jump points. You could certainly argue that the mere existence of the Terran planet in that otherwise empty system is what makes that system special. But even the most novice Starsector player would know that such a system is seriously boring compared to the likes of Corvus or Askonia or even Canaan.

And since building colonies seems to be the endgame up to now, wouldn't it make sense that the endgame content has a bit more sparkle than 1 star + 1 planet in some cases?

(To clarify, I'm not saying the procedural generator is bad. Far from it. It's one of the things that make Starsector fresh each time I play.)
Logged

Alex

  • Administrator
  • Admiral
  • *****
  • Posts: 23987
    • View Profile
Re: Procedural Generation needs improvement
« Reply #11 on: January 04, 2019, 10:11:06 PM »

However, if we acknowledge that blue star systems are mainly going to be used for exploration, whereas the player will likely jump at the opportunity to colonize a terran planet wherever practical and possible, then from a pure gameplay perspective, those systems in which the player invests the most time and resources building up should absolutely be more visually interesting, more expansive and content-rich.

Very good point. I'll give this some thought - maybe not for the .1 release, but I'll definitely keep this in mind in general as an axis that could be used to potentially spice up the lategame.
« Last Edit: January 04, 2019, 10:13:00 PM by Alex »
Logged

Sendrien

  • Captain
  • ****
  • Posts: 265
    • View Profile
Re: Procedural Generation needs improvement
« Reply #12 on: January 04, 2019, 10:24:24 PM »

That is all I can ask for. Thank you!

By the way, I definitely am not suggesting that Terran worlds should somehow be handcrafted. Perhaps one possible approach would be to give systems containing a Terran planet a "minimum level" of detail, e.g. at least 15 features, 3-8 of which are planetary bodies. Or something similar.

Given how elegant and refined your solutions to the economy have been, I trust you will find something equally impressive to spruce up special star systems.  ;D
Logged

Narvi

  • Ensign
  • *
  • Posts: 5
    • View Profile
Re: Procedural Generation needs improvement
« Reply #13 on: January 04, 2019, 11:47:42 PM »

I have a system with three 125% hazard rating worlds...

It was very special and felt good, but there didn't seem to be much benefit to having three colonies in the same system, which diluted it a bit.

Right now in my very late game I base system quality for colonization off whether there's an original comms relay or stable hyperpoint for comm relay. I don't know, maybe something more flavorful like having a ring of nebulae around the system give it a protection bonus would be nice.
Logged

Ishman

  • Captain
  • ****
  • Posts: 269
    • View Profile
Re: Procedural Generation needs improvement
« Reply #14 on: January 05, 2019, 07:23:06 AM »

Very good point. I'll give this some thought - maybe not for the .1 release, but I'll definitely keep this in mind in general as an axis that could be used to potentially spice up the lategame.

In my opinion, if you'd like to keep to the "habitable planets are rare" archetype of sci-fi setting, then perhaps any system containing a Terran planet would have one for a reason - be it a massive terraforming effort, or a particularly lush system. Speaking of lush systems - rocky planetoids that actually have just the right amount of water to only partially cover the surface seem to be an extreme rarity in the more sensitive exoplanet scanning results of late. It seems they're MUCH more common as moons to saturn/super jupiter sized bodies.

It'd be nice to see some more systems with desirable moons clustered in a habitable zone, whereas any of the nice planets would be likely to be the result of geoengineering, so in warning beacon systems or suchlike. Lots of other ways they could be spiced up, but I personally am always super tickled by the 'Ancient Relic' archetype, so I'd love to see more of that intentionally crafted in the procgen.

I think another avenue of adding interest to systems would be 'system-wide industries' that can be constructed once you've staked claim with a colony. Constructing asteroid mining operations, Belt Habitats, dropping Sun-Skimmers into the central star - some of the large scale megaengineering the domain was doing that's fallen into disuse under the splintered factional strife of the current Sector.
Logged
Pages: [1] 2 3