New Planet Tech is KILLER of Exploration (all terrain is tiling/repeating/not procedural/random)

[Agony_Aunt]

If you think about it, no it wasn't what made ED great in the past. There was plenty of repetition and more importantly, plenty of wholly unrealistic landscape in the old tech. Why don't you fly around and look for repetition in that as well? For ease, follow the texture tiling in the landscapes.

This crusade of yours seem to miss something: A control group.

S

I never felt like there was repetition before. Every planet, while being somewhat samey when of same type, at least had totally unique features. Every crater, every ravine, was totally unique.

I have no idea why they switched to this tile based approach, and can only presume it was something they felt would be better for the game in some way.... which feels like a mistake to me and now something they can't go back on.

 

[Erei]

I never felt like there was repetition before. Every planet, while being somewhat samey when of same type, at least had totally unique features. Every crater, every ravine, was totally unique.

I have no idea why they switched to this tile based approach, and can only presume it was something they felt would be better for the game in some way.... which feels like a mistake to me and now something they can't go back on.

There was repetition. But it was well hidden and you wouldn't see it. Like that stupid giant crater which is literally on every rock moon and planet. It was there in horizon to, but I didn't noticed it, because it was usually covered by the surrounding terrain and it "looked" different.
Now, this crater can't be unseen.

 

[Sysmon]

I never felt like there was repetition before. Every planet, while being somewhat samey when of same type, at least had totally unique features. Every crater, every ravine, was totally unique.

I have no idea why they switched to this tile based approach, and can only presume it was something they felt would be better for the game in some way.... which feels like a mistake to me and now something they can't go back on.

Well, with atmosphere comes current patterns and sedimentological features. Landscapes become more complex when winds are blowing. needed graphically distinctive process-based features added to them. Or we would truly just have coloured skyboxes to show for the update.

This way of using objects also allows for added variability for future expansions, when we get running water, clouds, precipitation, etc.

But with this first step into building landscapes in atmospheres, the components are few and we get repetition more obviously.

S

 

[Sakashiro]

There was plenty of repetition and more importantly, plenty of wholly unrealistic landscape in the old tech. Why don't you fly around and look for repetition in that as well? For ease, follow the texture tiling in the landscapes.

That's not how it works. You make the claim, burden of proof is on you.

 

[[VR] Merkir]

I landed on Kremavi A 2...I scanned there, and it looked like no patterns were repeated but I could be wrong, as I didnt go around the whole planet.

So as promised, for Kremavi A 2, here's what I found. I've shown the raw screenshots first, then spoilers of some of the repeating terrain. There are far more repeats than I could show, ofc. The first few were from an altitude of 20-50 km.

Spoiler

Spoiler

Spoiler

Spoiler

Spoiler

Note that for some planets it's difficult to see repeats from high orbit. This seems to occur for planets with many terrain/biome layers. That's why the heatmap is very useful, as it often highlights only some of those layers, as shown in the final image.

 

[Ettanin]

Randomness contradicts the fundamental principle of a turing machine: Determinism.

You can't get true randomness with determinism because randomness is based on chaos and chaos is not deterministic.

A good pseudorandom generator hides its deterministic roots. This terrain generator lacks the amounts of samples to do so.

Horizons had a much higher sample rate thanks to its vertex based generation.

 

[Sysmon]

That's not how it works. You make the claim, burden of proof is on you.

Nah. I don’t see repetition of any significance. Can’t prove a negative.

Besides, I’m more interested in exploring the wonders Stellar Forge has made for us instead of looking for perceived failings of it.

S

 

[Sysmon]

Randomness contradicts the fundamental principle of a turing machine: Determinism.

You can't get true randomness with determinism because randomness is based on chaos and chaos is not deterministic.

A good pseudorandom generator hides its deterministic roots. This terrain generator lacks the amounts of samples to do so.

Horizons had a much higher sample rate thanks to its vertex based generation.

Nature, which the game tries to simulate, is not random.

S

 

[jojon]

A coastline will have any amount of bays, but never the exact same bay occurring over and over again.

 

[Ronzer]

So as promised, for Kremavi A 2, here's what I found. I've shown the raw screenshots first, then spoilers of some of the repeating terrain. There are far more repeats than I could show, ofc. The first few were from an altitude of 20-50 km.

View attachment 248110

Spoiler

View attachment 248111

View attachment 248112

Spoiler

View attachment 248113

View attachment 248114

Spoiler

View attachment 248115

View attachment 248116

Spoiler

View attachment 248117

View attachment 248118

Spoiler

View attachment 248119

Note that for some planets it's difficult to see repeats from high orbit. This seems to occur for planets with many terrain/biome layers. That's why the heatmap is very useful, as it often highlights only some of those layers, as shown in the final image.

Ahh, thanks for that! So many uses of it, instead of only a few spots.
So the solution would be the image it generates for the planet, being sliced up, but if any area of it is used, it cant be reused, whereas now, it slices it up, but randomises how the sliced up parts are used/reused.
If it breaks things for them to redo this, I'd be ok if they didnt. But if they were able to, or it is bugged, as they suggest, then it'll only be better in future when they fix those bugs.
Keep at it FDev!

 

[Drelthar]

Frontier can't find a solution - but the solution is obvious, it would just take a LOT of work, and require "re rolling" the planets.

Make a few hundred thousand (not exaggerating) new assets to use for the proc gen - then create a system that can procedurally modify those assets here and there, then block the same asset from appearing the same -system- (preferably galactic region but that would require more than a few hundred thousand more assets)

This would reduce the starkness of the rabbit inside the wizards hat and be a more convincing illusion.

Alternately work REALLY REALLY HARD on the system that can modify assets to make it exceedingly good at mdifying the assets, so they keep most of the details, but change it enough for humans to struggle to see the similarity (this probably requires a deep learning AI though....) to avoid having to make so many new assets.

 

[[VR] Merkir]

Frontier can't find a solution - but the solution is obvious, it would just take a LOT of work, and require "re rolling" the planets.

Make a few hundred thousand (not exaggerating) new assets to use for the proc gen - then create a system that can procedurally modify those assets here and there, then block the same asset from appearing the same -system- (preferably galactic region but that would require more than a few hundred thousand more assets)

This would reduce the starkness of the rabbit inside the wizards hat and be a more convincing illusion.

Alternately work REALLY REALLY HARD on the system that can modify assets to make it exceedingly good at mdifying the assets, so they keep most of the details, but change it enough for humans to struggle to see the similarity (this probably requires a deep learning AI though....) to avoid having to make so many new assets.

That would probably require more assets than you think. Many of the repeating "tiles" are around 1 km² (1km x 1km) in size. An earth sized planet would need about 500 million unique tiles. Although as you say, procedurally modifying a smaller subset could work, as long as the transforms work well.

 

[Pigletslastand]

The clip of Braben from 2012 said that the one thing wrong with fleshing out proc gen with hand made assets is if you do it with too few assets it becomes noticeable. He actually PREDICTED this very problem. So yeah when the head honcho flags a prob and that prob makes it into the game. Direction has been lost somewhere along the years!

 

[Drelthar]

The clip of Braben from 2012 said that the one thing wrong with fleshing out proc gen with hand made assets is if you do it with too few assets it becomes noticeable. He actually PREDICTED this very problem. So yeah when the head honcho flags a prob and that prob makes it into the game. Direction has been lost somewhere along the years!

That would probably require more assets than you think. Many of the repeating "tiles" are around 1 km² (1km x 1km) in size. An earth sized planet would need about 500 million unique tiles. Although as you say, procedurally modifying a smaller subset could work, as long as the transforms work well.

Yeeeeep.

I am "fairly sure" that Horizons used some pre made assets, however, it had a lot of layering over the top with "true" proc gen. This masked the issue pretty damn effectively.
I'm not sure they can do this with the new system though, as it's already taking an awful lot of time/power to generate from the assets it's using, so adding another layer on top (they must already have some just not the same extent as horizons did) might cause some real problems performance wise.

 

[Agony_Aunt]

There was repetition. But it was well hidden and you wouldn't see it. Like that stupid giant crater which is literally on every rock moon and planet. It was there in horizon to, but I didn't noticed it, because it was usually covered by the surrounding terrain and it "looked" different.
Now, this crater can't be unseen.

That's what i mean, every base terrain feature was deformed by various other features to make each thing unique.

 

[Agony_Aunt]

Well, with atmosphere comes current patterns and sedimentological features. Landscapes become more complex when winds are blowing. needed graphically distinctive process-based features added to them. Or we would truly just have coloured skyboxes to show for the update.

This way of using objects also allows for added variability for future expansions, when we get running water, clouds, precipitation, etc.

But with this first step into building landscapes in atmospheres, the components are few and we get repetition more obviously.

S

Looking at what we have now, i feel as though just a coloured skybox would have been the better option, and delayed any surface changes until they were sure they had a good solution. Because as things stand, FD's seem to be saying they don't know how to make the current system better.

 

[Drelthar]

Looking at what we have now, i feel as though just a coloured skybox would have been the better option, and delayed any surface changes until they were sure they had a good solution. Because as things stand, FD's seem to be saying they don't know how to make the current system better.

They were sure they had a good system though.

They were just... unfortunately, a tad wrong about that, because they misjudged how well the player base would be able to see through the illusion (it would always, always be an illusion no matter the system)

This does seem odd though, considering they must know what their community is like.

 

[Pigletslastand]

Could perhaps the problem be that any changes at this 'base level' be a real headache when they plan to expand on it in the future? We know from Dr. Ross that they are currently working on the next iteration on planets so there is more to come and if the whole team were working on a second tier of detail only to be told the foundation was wrong this could really put a spanner in the works and may explain why it's taking so long as they don't want to ruin all the work they've evidently moved on to.

This is all conjecture and without clear dev explanation almost meaningless but if they do decide to explain they run the risk of losing the reveal. However look at CIG, if their business model proves anything it proves that players really REALLY are willing to hang on and give you the benefit of the doubt if you show them what you want to do. I'm not advocating it though. The problem with the converse is that you need a lot of faith and goodwill, and Fdev have squandered theirs.

 

[Pigletslastand]

Here's the problem.. They are too self assured in their own predictions without testing them on a real player base.

They were sure they had a good system though.

Remember Dav (bless his heart, really like him) confidently announcing on stream that they had run the numbers and were correct on the pricing of Fleet Carrier upkeep?? Yup, they see through spreadsheets, not player experience alas.

 

[AkenBosch]

Looking at what we have now, i feel as though just a coloured skybox would have been the better option, and delayed any surface changes until they were sure they had a good solution. Because as things stand, FD's seem to be saying they don't know how to make the current system better.

This. I can't but feel that the current tech suffered a "wind tunnel disconnect", that is, like those car concepts with revolutionary ideas and aerodynamics that work wonders in the engineers simulations and wind tunnel tests, only to reveal a terrible lack of correlation with real data when confronted with the harsh reality of a road test. I can speculate (talking out of my arx here as usual, so bear with me) the new planetary tech might indeed have been a great concept on the drawing board and with great (probably necessary) potential for going forward with future developments, but once put "on the road", the compromises required to make it viable (read: ready on time for release with adequate performance) were harsh and not at all foreseen. The "ultraforcapture" setting stays there as a huge waving yellow flag....

Anyway, as mentioned, back at the time of their first implementation, SC was mocked to high heavens for the blatant tiling visible on its handful of planets, and now I seem to remember they look just fine, as in, they are probably still tiled as before but the thing is masked a lot better. I guess ("/hope/pray/no way", pick your choice depending on the amount of salt in your diet) they'll find pathways to sort that out with an acceptable compromise in the future, eventually.