Observing choices does not really tell you much about the player.
Contrary, if the game has no choice at all, it's the player doing his choice while selecting game A over games B,C,D. The player may still regret his choice, though.
You are right about the reality of the player making choices and not liking the outcome. Might be that they thought something else would happen. And they may even try again and just do something slightly different to try to get the outcome they were looking for.
A story usually has ups and downs. Having events that feel awful is often part of the story and even if they don’t like it in the moment, they can like the experience in the end, having gone through the bad experience.
I believe that games can get progressively better at understanding what the player prefers and why, and that systemic games make this possible.
The idea about having a game that adapts is based on that you should not force every user to configure the game before they play it. This relates to how many games start by asking the player if they want the game to be easy, medium or hard. They may also ask other questions. It's good that the option to adapt the experience exists. But there are ways to do it inside the game.
It’s not a super-complex concept to implement. You can do it with an in-world difficulty setting such as a ring of invulnerability. A fork in the road with one road leading to rainbows and bunnies and the other to hellfire and demons. The idea is to create some form of first-time default experience that should work for most players, but provide optional activities.
Adaptive difficulty can often be done by a wide range of success possibilities. A simplified example of this is getting to the goal, but having three stars on the way that you can collect if you want, but getting all three is extremely difficult. The player will set their own goal of what they consider to be a success.
If quests have the option to solve them by talking, sneaking or combat, the game should have a pretty good idea of the playstyle after the tenth quest. Especially if how easy each of the pats are varies from quest to quest. There may be one where the combat path is super-hard but the talking option is super-easy. – There should probably be some sort of tutorial part for players that may not even have realized that you actually can talk to enemies.
The option for having a guided experience can be presented in game as skills / upgrades / widgets that you can choose to use or not. If you have a map that the player looks at every 10 seconds, they may want it on screen. Or they may appreciate in-world markers. I usually prefer to play with no HUD, but want the option to bring up in-world markers to orient myself on demand.
The feeling of how long a main quest is can be adapted. The length of a story is created by the old traditional story technique of nesting and connecting stories. But this can be adjusted mid-game. The player may say that they feel that the story should start to wrap up, the game can adapt by not introducing even more sub-stories and complications. All this can always be adjusted explicitly by going to the game config menu and telling the game to wrap up. But it can also be done automatically by measuring changes in how much and for how long the player engages with the threads of the story. It could probably even detect if the player is about to fall asleep.
The game can detect how much the player cares about people getting hurt. Some people with PTSD have a hard time playing most games since the stories are incredibly brutal. Creating the best game in the world for these people, is not a high bar, since there actually doesn't exist AAA action adventure games without brutal tragedy. The game can detect their feelings about grueling setbacks and hurting NPCs by giving lots of options to help NPCs hurting, and see if they go for the rewards or for the selfless acts. If they chose to make friends with everyone instead of knocking them out and taking their stuff, the game can adapt and avoid having the murder of your best friend a part of the story.
A systemic story game will have some systems implemented for supporting certain types of stories. Expanding it for more types of stories and situations is sort of a never-ending project. But supposing that we have implemented a couple of genres and several themes and topics, the games should be able to see which of them that the player engages with. It may be because of personality, environment, mystery or something else. But over time, it should get an idea of the types of topics that interest you.
The themes come with moral dilemmas. When they are starting to be implemented in a way where you actually have a relationship with the people and have the option to explore life philosophies. The NPCs could straight out ask you what you thought about if the outcome was good, what you would do differently, if you regret something and so on. These are the types of things where the game can learn more about your soul.
And even while doing X, you may still almost certainly fail to give me the exact choice that i would want to do. So for me not much seems to change. I still have to pick the choice i dislike the least, but you have to work much harder on making the choices work.
Thinking of it, this is my main concern:
So for me not much seems to change. I still have to pick the choice i dislike the least
Can you give me an example of what choice you are missing that could be missing in a systemic story game?
Having feedback is important. And as I mentioned, I will default to give as complete insight as possible, until the point that uncertain feedback can be done in a good way and the player has selected to play the game with that uncertainty where the avatar may get better over time at perceiving and intuit the result of their actions.
It should be clearer than “Erik will remember that”. I’m planning for multi-dimensional relationships, so it won't be a “+1 relationship”, rather, as feedback to what you just said could be presented with something like; “Erik is now a bit afraid of your unpredictability and will hesitate to trust you with something important”.
Could be as simple as letting the player die often, so he can select different choices next time.
Setbacks are an important part of many stories. Games as they exist today, have trained players to not accept sub-optimal outcomes and will willingly die or reload a previous save if things didn’t go as good as they were aiming for. That is why setbacks and failures often are relegated to cut-scenes. Sid Meyer commented on this, and how he had to abandon the idea of creating awesome stories of comebacks, since the player would simply reload and try again. This is why permadeath is even more important for systemic story games. Rewind should be an option, but not the default. Many players don't probably understand or accept that losing in combat is a vital part of a good story.
It should also be a part of the game design to reconsider choices and take other paths. This is not a branching story. You are free to go back to any NPC and say that you have changed your mind. You can correct misunderstandings. You can fix your mistakes. Failure is not the end. Even death is not the end.
Ideally, the game should be continually saved as an event log so that you can go back to any point in time to try out other things. You should also be able to share those scenarios and see how other players handle the situation. But the main and default game must be that you live with your choices and let them play out.
aigan said: I think Aleissia Laidackers description was pretty good, by the use of matrices for seeing how interconnected the systems are. That was the thing I was talking about in relation to NetHack.
I remember your NetHack blog post. The story about having hands or not, combining stuff, etc.
I can tell a similar story from Super Mario: My wife and i play it daily in Splitscreen, since 10 years i guess. Actually a PC clone with levels made from its community.
Within the last year, my wife has figured out something new. You might know, if you ride a dino, and you eat a blue shell, you can fly. But you can also do this: Jump off the dino, jump on a yellow turtle. Don't kill the turtle and don't get killed from it. Wait until it gets back in its shell. It's a precise and difficult operation, usually. Now, back in it's shell, it becomes a rainbow shell, because it was yellow. That's dangerous, because it's one of a few enemies which have acceleration, and it chases you. But you have already mounted your dino again, and you are ready to eat the shell. When you eat a rainbow shell, you can fly too! No need for a rare blue shell. : )
Super Mario is full of such stuff. Is it a systemic game? Yes, it surely is. But there are no clumsy abstractions, no complexity bloat, no interruptions, it's all done within it's laws of physics and at real time, within it's usual controls, during moment to moment gameplay. There is not even a inventory screen. If you want to consume the powerup from your inventory, you press a button and the powerup falls down from your single top screen inventory slot, and you have to catch it. If you miss it, bad luck.
This game is pure perfection, and it did everything 100% right, long before we came up with fancy terms to describe how things should be done eventually. The philosophy may be useful to communicate, but it is not necessary, nor is it something new.
aigan said: Using physics simulation would be nice, but would that include the individual candles and the situations in which they will blow out or what would happen if an arrow hits it?
Yes, absolutely possible. I can make joints so that they break on a certain amount of stress, and i could make candles track velocity to blow out. After that, the chain reaction could be seen as ‘truly emergent’ in my terms. But ofc. i will likely implement those things only after i came up with the idea of the potential chain reaction. Then it's no longer emergent, but just intended by design. A matter of philosophy or semantics i would say. But if it works, that's all we need. In practice, playing around with simulation often does show emergent behavior. It may inspire us to use it by design. Same thing as improvising on an instrument, coming up with a catchy melody by random luck, turning it into a song. And afaik, all great songs were written this way. It's not possible to do it by intent. In gamedev it is much harder to improvise freely, but it is possible to some degree. And maybe it's the best way to come up with good ideas also for us.
aigan said: You could also determine the sound produced by the impact from the floor material, the chandelier and the impact damage.
Yes. Physics engines have contact callback to implement things like that. Sounds, particles, but also things like conveyor belts can be / are done this way.
However, our options regarding fluid simulations are quite limited. Usually we can do it only inside some box to bound costs. Wind is often modeled just procedurally for graphics. Yo can see that some trees wobble in the wind, and some other plant right beside does not move at all or in a completely different direction. Water usually has some fake physics going on only at the surface. Don't expect to change this anytime soon. Focus on rigid bodies.
If you're interested, HZDs physics engine ‘Jolt’ is open source on gitHub. I've tried it recently, and could compile and try the demos without issues. It's smooth and fast, but i guess too springy for my robotics needs (i use Newton). It has nice soft bodies, raycast cars, ragdolls with animation blending, probably a player controller as well. Standard features. If you want to prototype 3D game ideas, a physics engine gives you all you need. Just no animation or fancy rendering. For that you'd need a game engine.
aigan said: That will make all the systems and tools relevant for progressing in the game. All the different elemental arrows, canisters, traps, potions, and the intricacies of the machine's behavior.
My problem here is always the same: I want to explore the options, but i can not remember them. If i come back after a week, i have forgotten it all. It's not intuitive enough, and so it remains hidden from me in practice.
So it is not bad that they make it optional, and a dumb play style works as well. But it would be much nicer if i could just figure out options on my own, because related interfaces are intuitive, functionality is obvious, etc.
I'm actually happy to be so dumb and lazy as a player. I think this can make me a better game designer eventually.
aigan said: I can play the game and it can still be sold to others who don’t care about pacifism.
Hehehe, it's just NPCs, dude. Shoot them an eye out, let their blood spill, and be nice to the real people instead. \:D/
aigan said: Immersive sim games are probably not for you then. Or it could be that you didn’t play it on a difficulty high enough to force you to find new approaches. Many players don’t like standing still for minutes while observing the situation and thinking out plans. That’s ok. There are other games they can play.
No. ‘Immersive sim’ totally sounds like the thing i want. The RPG promise also sounds like i should like it. And i have no problem with observing some situation, working on some clever plan on how to deal with it then. But i don't understand those games, conclude they do it wrong, and only because of that i play Quake again.
aigan said: Either way. The home computers/consoles have had the capacity for the last 20 years or more. But everything went to push pixels.
It's just that dear game developers had - and mostly still have - no idea what to do with insane GPU compute power. They focus too much on brute force pixel shaders. The PS4 was never maxed out even slightly. It can do this path tracing nonsense you see now on 4090$ quite easily. At least if you implement a proper technique which is actually meant for realtime, uses GPGPU as intended, and accepting some lack of high frequency details. It really sucks i still work on the necessary tools, although i have the realtime GI stuff working since 10 years. I could have helped a lot with saving power and costs wasted on shiny pixels. But well, that's another story. However, a cheap APU with a 4tf GPU is enough for games. Any other claim is based on incompetence or greed, and hopefully i can proof it soon enough.
aigan said: My mentioning about 2035 was just for the length of implementation.
Maybe a realisitc estimate. I always thought i need to multiply my estimate with a two, to get a better estimate. But then it became a five. At least, for now…. : )
aigan said: There is a reason for generating the world instead of having it static. The hero pushes forward, wounded and getting ever more close to death, certain she will find the hidden safehouse. Now, at the moment of last consciousness, she finally reached the door. And, with the ability to change the world, that door could be something that would fit the story, instead of just the laundromat that was placed there as a static world location.
How awesome. A specific door, tinted in blue to express the sadness of the current situation, with cracks on the walls to express how things tend to die after some time.
Made by whom? An artificial computer program replacement for a proper artist? Oh, i forgot - computer programs no longer exist. They are currently replaced themselves with ML models.
No thanks. I prefer the static stuff, made by an actual human with a soul. It may not always fit the situation, just like a real world lighthouse wouldn't. But it will still work, and feel better overall.
There are other ways to have dynamic atmosphere. Weather for example. ; )
aigan said: Every bit of content that’s fixed is just another rope that ties the storyteller's hands.
No. The real storyteller tunes the content before he tries to sell the book.
You want to sell a book that is empty, reads the readers mind, filling the page with text as the reader reads it. You may succeed, but this does not make you a storyteller. You'll get a handshake from Geoff, but no Nebula price.
aigan said: If you can have unique textures everywhere, you should be able to have uniquely generated things everywhere.
Yes, that's the plan. Storage space is the limit. And it's a very tight limit, since for production i need much more than for the shipped data. Working on related compression ideas is still in front of me.
aigan said: but I don’t want to see a thing I should be able to use but can’t because it's just dead geometry. If I see a branch on the ground, I should be able to pick it up and use it to reach something.
Yeah, but there is no way around compromises here. We'll never achieve total realism. Just think about sound synthesis. Cracking the branch should create a unique sound, coming from the simulation of cracking wood. We will never get there.
But that's good. It means we can always strive for more, constantly achieving progress. It's basically guaranteed we'll never get stuck. This gives us a serious advantage over other creative industries. We're lucky much more than restricted.
You should be able to carve your initials in the rock. And since you use unique textures, that shouldn’t be a problem.
Yeah, that's possible and needed for decals, footsteps, displaced snow, etc. But it's only a visual effect and should not affect physics and gameplay. It's not worth the cost, except games like Mud Runner. In theory we could even simulate 2D water on such surfaces, eroding sand and such stuff, at very small scales. But usually there are better ways to invest your performance budget. And such deformations can't be persistent without exploding player hard disks.
aigan said: Would very much like to have better terrain generation that includes geological processes like landslides, erosion and more.
I'm working on this. This is probably the first terrain simulation you see which is entirely 3D instead heightmaps as usual:
There are also Aresseras blog posts here, he generates entire planets including erosion simulation in realtime, which i've also never seen done before.
Procedural generation is fun to work on, contrary to tedious geometry processing. So i hope i can continue on this somewhat soon… : )
However, commercial terrain tools are pretty damn good. It's just that when you zoom in to ground level, heightmaps become low res, and procedural texturing lacks variety and features. So the end result we see in games isn't as impressive as looking at the whole model from large distance. Storage limitations.
aigan said: From my perspective, the computer should have lots of time to work on simulations and generations since for the type of game I envision, I would sit around with my virtual friends, in the same spot, and not demand very much from the 3D engine.
I'll saturate background processing capacity alone with decompression and generating some highest detail levels using texture synthesis. Generating all the unique stuff would be impossible for me. But other engines can generate at realtime using heightmaps and procedural placement of instances, which is currently state of the art everywhere.
However, sitting still and doing nothing does not free up rendering resources. They recompute all the pixels every frame, no matter if they change or not. TAA, upscaling and frame interpolation are somehow an early attempt to address this. But we are far from rendering which is really efficient, doing only what's needed. The entire field is dominated by a brute force mindset. When GPUs were introduced, brute force was the only way for a long time. It's the standard, still working well enough. It made Jensen very happy, until the rise of ChatGPT made him even more happy. ;D
aigan said: But why not use the cloud also? When you could take even more time to simulate thousands of years of subtle changes.
Cloud is the only way to solve the storage problem. But if we do this, it will take many mega watt hours to transfer the data. It may be the future. but like AI, it is a future i do not want. I'm fine with less content, less details, less of everything. I'll be fine with a power efficient mini PC pretty soon. No more dGPU. I'm modest, and have learned to lower my expectations. Which is the only way. Either that, or we will all die. (It's already pretty hot here in central Europe)
JoeJ said: However, commercial terrain tools are pretty damn good. It's just that when you zoom in to ground level, heightmaps become low res, and procedural texturing lacks variety and features. So the end result we see in games isn't as impressive as looking at the whole model from large distance. Storage limitations.
This is the problem with offline procedural generation, you need to store it somewhere. Consider a game world area of just 32x32km (the most you can do with 32-bit float, and reasonable by today's standards), where you want to generate terrain down to 0.1 meter resolution. Even if you only store a 16-bit heightmap, you would need 320,000 x 320,000 x 2 = 190 GB of space. Just for the heightmap, not including textures etc. On an earth-sized planet you would need 10^17 bytes (89 petabytes, enough for big data center).
This is the big differentiator of runtime procedural generation - it allows for massive worlds with nanite-level of detail using only a few GB of memory. There are no issues with streaming that much data from disk. In my system I can fully generate terrain to 0.1 meter resolution on an earth-sized planet in less than 5 seconds on an 11-year-old CPU (using 4 threads). The main bottleneck is uploading vertices to VRAM. At high detail it becomes a significant overhead to upload vertices and keep consistent frame rate while moving quickly. Since I added asynchronous generation recently, it's possible to fly across the planet at many km per second with only a 1 second delay for the terrain to reach target LOD subdivision level (i.e. almost no pop-in).
I'm aiming for quality similar to offline erosion tools, but at runtime. There will be some limitations due to inability to do lots of erosion iterations, and any kind of 3D erosion like you are doing is not possible, but I believe I will be able to produce very nice realistic results sometime this year. Hopefully I can hack in some voxels/marching cubes at the end of the erosion pipeline to generate nicer looking cliffs. After that I will start work on an actual game using this tech, which is planned to be a “hard sci-fi” survival sandbox game, with a focus on realism. I feel like this sort of first person game (similar to The Long Dark, Green Hell, Subnautica but on the scale of No Man's Sky (without that game's awful design)) is possible to finish by myself with some hired artist help. The environment is the biggest part of a wilderness survival game, which I already have significant progress on.
aigan said: The themes come with moral dilemmas. When they are starting to be implemented in a way where you actually have a relationship with the people and have the option to explore life philosophies. The NPCs could straight out ask you what you thought about if the outcome was good, what you would do differently, if you regret something and so on. These are the types of things where the game can learn more about your soul.
This one sounds interesting. The other options above in the post feel like too much customization for me. It's a bit like attempting to serve the player at any cost, not being ashamed about looking submissive and desperate. ‘Dear player, my game sucks. But maybe you can make it a bit enjoyable by adjusting all those knobs i'll show you on the next page, to your liking.’
aigan said: I believe that games can get progressively better at understanding what the player prefers and why
Cool. I'll use it to torture the player even harder. >:D
aigan said: Can you give me an example of what choice you are missing that could be missing in a systemic story game?
Sounded like a difficult question at first. But then i thought - i like open ended stories. It shall raise questions, keeping me thinking for a while after it's over. Most people dislike open ends, but to me, a good story that is not open ended is not really possible.
So how would your player analysis figure this out?
Or, another example. Bad guy asks me: ‘I've got them, HAHAHA!!! Here, see, your wife and your daughter. Bound on the table. And here is my shiny new butchers knife. Make a choice! Which throat should i open in front of your eyes? Hers, or hers?’
I expect the following options:
“Take my wife, but i'll kill you!!!”
“She's pregnant… so, take my little daughter, but i'll kill you!!!”
“Take me for them! Let them go! I beg you, take MEEE!!!!”
But i don't want any of that. I want to say:
“pffff, idk. Take them both. Might restart to give them better hairstyles anyway. Go on and kill them. Can i go after that?”
Yeah, that's a good answer. :D
aigan said: But the main and default game must be that you live with your choices and let them play out.
Ok, i think i can accept this. Ideally it becomes noticeable that my current choices actually are a result of former choices, and that it is not just branches through a fixed graph, for other reasons. But i still think you need something like a key scene / system seller moment, as early as possible in the game. Maybe you do not need to inform the player directly about the new feature. It would be much more impressive if you could just show it with the early gameplay itself.
Aressera said: The environment is the biggest part of a wilderness survival game, which I already have significant progress on.
Then i guess you'll skip architecture, but you'll need alien creatures? Any plans on how to do this?
I never really made it past that point as a single dev. I've had animated walk and run cycles for a human model, which was a week of work just for the animation clips, but then i fell back to capsules for ‘simplicity’.
JoeJ said: Then i guess you'll skip architecture, but you'll need alien creatures? Any plans on how to do this? I never really made it past that point as a single dev. I've had animated walk and run cycles for a human model, which was a week of work just for the animation clips, but then i fell back to capsules for ‘simplicity’.
I plan on hiring artist(s) for that. We'll see how that goes… There is a lot of work to get ready to ingest that content. The biggest challenge for me is building the engine infrastructure to support animated characters. I have basic animation capabilities, but without an FBX importer it's hard to make much progress yet. It also seems like a lot of work to make a character controller that drives animations in a realistic and smooth way in response to player or AI inputs. These seem like the bottlenecks for my project at the moment. I also need to do tree/plant/boulder generation, which is required for finishing touches on terrain. It's certainly a lot of work.
Using Unity or Unreal I'd get a lot of that for free, but would be completely incapable of handling realistic-scale planets without tricks like my engine can do, due to its deep integration of multiple coordinate systems into graphics, physics.
So we either need to find ways to implement crafting in the game, not it's menu, Or we need to focus on other systems which build on top of the actual simulation. But so far, nobody could solve either of these problems well, although they are really obvious to see. So maybe it just isn't possible, or we are all blind since decades to see a potential which hides in blind sight.
You are right. Much of gathering and crafting is bad and needs to connect better with the world simulation.
I don’t like the collecting, inventory and crafting systems in a lot of games. It’s based on a type of progression hook where it’s just about getting more and better things, with cycles of upgrades. Much of it doesn’t have to exist in the first place.
Having to find food and water is interesting if it’s something new to learn or part of a challenge. It's relevant in the beginning when you learn about the world. After a while, it’s no longer something you should have to manage at all. That is, until the time the story takes a turn. At some point, you will find yourself in an unfamiliar hostile environment and have lost all your stuff. Now the food and water system is relevant again.
The metroidvania mechanic is better where you get or upgrade tools that give you new capabilities, but not if it's treated as a lock and key. It should be systemic, so that you could potentially think of five other ways to do the same thing without that tool.
My general point about skipping repetitive boring stuff is based on story structure more than how games are today. Movies usually have a lot of time-cuts. Especially if it’s an epic adventure. The game version of this can have a lot of repetition if it’s in service of mastery, building on a tone, or in other ways are enjoyable. But there will still be a lot of stuff that should be part of the story but not played out second by second. If it’s not a story game, you can design it in a way to eliminate those menus by creating things in the world that fill the needed function.
But I don’t like when the game goes the other way and puts menus in the world. There are many games with hubs where you will run between a couple of stations just to manage different aspects of your resources. Even worse when it’s about collecting or buying resources, you will have to run or fast-travel between parts of the world just to get the stuff you need, without making any interesting decisions or any type of challenge. If it’s traders you already visit or people you already spoke to, those should be accessible in menus, where you can go back to the memory of what people said to you or the memory of what different vendors buy and sell and to what price. You should be able to work out all the things you need, and then done, you can do the equivalent to the commit action, and only when will the game simulate the actual events of visiting the needed traders and resource stockpiles, with the passing of time and things that might happen on the way. That way, you don’t have to run back and forth just because you forgot who it was that sold that specific item.
Horizon along with several other games has a campfire mechanic that will skip the whole upkeep phase of repairing equipment, eating and sleeping in just one click. If there are more choices that shouldn’t be played out in realtime, it’s back to menus. Some resource management games, such as colony builders have ways to set up resource goals. This is for automating the task to retrieve and produce the amount you decided you want. In an open-world adventure game, this can be done by dealing with hostile actors in an area (befriending, driving away, capturing or killing) and setting up camp. At a camp in a safe area, you should be able with a single action to gather all the resources you need from the area, including hunting for meat, gathering herbs, and filling up water and so on. You will take care of wounds and prepare for future healing needs. You will take care of your equipment, produce ammunition such as arrows, produce food for eating and for having during the next few days. And any other activities, like sewing new clothes or anything else. Based on the time you arrive at camp and when you need to sleep, you will start with the base needs and use the rest of the available time for things that take longer time, like the production of new stuff. At the start of the game, or for new things, it can be a manual in-world activity. And all of it should be tied to the world simulation. But anything repetitive can be set up as goals and be time-skipped.
But if you get it right, i would not be annoyed about your system, like i'm about crafting, collecting resources, or leveling up RPG characters.
The whole focus on skill trees and getting better weapons is just another version of the progression hook. I don’t like any of these number games. Having a fancy looking armor with a lot of stats that has no connection to the actual material and design of the physical thing is just stupid. That’s another topic, but I would like it if it was more consistent. If the player can get that much better in just a couple of days, her enemy should be insanely powerful. The systems should be the same for everyone.
A systemic story game should be good enough without those progression hooks. They are usually not part of stories in other media. The ideal design would be if the player can take on bigger challenges by getting better at the game rather than stat buffs.
So we need new and aggressive ideology and ideas. Star Wars, Bladerunner, Lord of the Rings and Dune - it's all gone. We need new stuff, no longer the SciFi helmets and trapezoid doors, swords and pointy ears, which look the same in every game.
Scorn looked interesting, but I wouldn’t want to play in a dark or depressing world. The topic of organic technology is interesting and has some fascinating themes. I like Farscape.
But nothing is new. What is it you want to explore?
More generally, I think its about the culture of reference. Each cultural product is building on the thought patterns built up from previous successful products. At some point, we need to re-establish connection with the human experience, not seen through the cultural kaleidoscope.