Libre Speedrunning

[drummyfish]

Is there any libre speedrunning community? Both libre gaming and speedrunning are a niché communities, and I imagine their intersection will be super small, but it could still be fun.

I know about a few libre games that are a subject to speedrunning -- the latest one I've seen being Marble Marcher. Then there is XMoto and Neverball, which are completely designed to suit this idea. I think I've also seen SuperTux TAS and STK speedruns around here on the forums.

I think libre games, thanks to being so open, could be especially fit for the TAS runs. In proprietary gaming, these are mostly limited to games played on emulators, but with libre games you can do all kinds of crazy stuff with interfacing the code to your tools, or simply modifying the game directly to allow for recording an ideal replay that can then be played back in an official version.

Personally I have never done a speedrun, but I've been dedicating a lot of time to Xonotic lately, getting some skill, and I have gotten an idea to speedrun the singleplayer campaign -- it's not a completely predictable game and every run would be different... though some levels have tricks to complete them very quickly -- it is interesting to explore this game. For start I'd like to do a segmented run, then later perhaps a single segment run.

Ideally of course the libre speedrunning community should share the records and videos of the runs under free licenses as well. I intend to do it with my Xonotic run.

Are there any other interesting games that are subject to speedrunning, or potentially would be? Is there anything more to say about this topic? Is the speedrunning community -- which is arguably bigger -- aware of libre games in some sense?

[acme_pjz]

You can try Me & My Shadow https://forum.freegamedev.net/viewforum.php?f=48 which is a puzzle platform game with replay support.

[drummyfish]

Me & My Shadow -- thanks, that's the kind of thing I'm looking for!

As I am thinking about it, a libre speedrunning community to me would imply something more than just speedrunners who play games that happen to be free. The surplus value that speedrunning adds to gaming should be based on free society principles here -- so as I said, the videos would be free, hosted on a free platform, the discussions would be on free forums, the tools used for TAS runs would be free and shared among the members... and so on.

[onpon4]

I think ReTux is very speedrunnable. One time someone was going to do some sort of TAS, but they were making all this custom software and never actually got around to finishing it to my knowledge.

Seeing a TAS of ReTux would really make my day.

Project: Starfighter might also be a TASable game via luck manipulation. I imagine a game where luck manipulation is used to get super charges every time and blow through the levels like crazy. The fact that the game uses a mouse might make a TAS tougher on the technical side, though.

Ooh, Torus Trooper! Seeing that game going at a speed turned up to 11 with a TAS would be awesome.

That's about all I can think of.

[drummyfish]

Seeing a TAS of ReTux would really make my day.

I will keep this in mind I actually haven't played it yet, gonna give it a try. Thanks for the other recommendations as well.

[drummyfish]

Here is my first Xonotic run -- segmented -- pretty terrible, but I'd be very glad if someone felt challenged to improve it and start a trend in speedrunning it:

https://peertube.mastodon.host/videos/w ... e58536da18

I've now downloaded reTux and giving it a go.

Also another pretty runnable game would be Freedoom.

[Julius]

Nice video (and yay for using peertube!), but since bot matches are so random it is really speed-running?

I am not really into that stuff, but it always seemed to me that it was about static environments or highly predicable enemies in Mario like platformers. As a result you see crazy optimized speed-runs for these games, which simply doesn't seem possible with random bot play.

[drummyfish]

and yay for using peertube!play.

Only free SW and free resources were involved in making that video, from bottom up, including the OS (completely free distro), video recording, editing (Blender), the timer (it's a CC0 file from Wikimedia Commons), the font and so on. The video itself is free too.

but since bot matches are so random it is really speed-running?

AFAIK non-predictable runs are perfectly okay, people run things like GTA and so on. But the predictable ones look much nicer of course because of the minimum errors, so I guess they're more common.

[drummyfish]

Alright, and now I've figured out something very neat (the command is "slowmo 0.1"):

https://streamable.com/z1onm

[Lyberta]

Deleted.

[onpon4]

As a result you see crazy optimized speed-runs for these games, which simply doesn't seem possible with random bot play.

In a game where randomness really makes a difference (Project: Starfighter is another example), the majority of the work going into the TAS is going to be luck manipulation: spawning enemies closer to you, getting lucky items, etc by controlling timing to manipulate the seed.

Of course, for that, you absolutely need to have a very precise and controlled environment, preferably an emulator if possible (though there are other options as well).

[Julius]

including the OS (completely free distro)

Wow, I didn't watch the video but, did you get any FPS? I don't know of any GPU that runs in free distros, when I tried Red Eclipse without blobs, I only got 3 fps on my 4 GHz 8 core CPU.

Both Intel and AMD GPUs have fast open-source drivers on Linux these days, unless you are referring to some firmware binary blobs that AFAIK are utilized for all GPUs on the market today.

Back on topic, "luck manipulation" reminds me of the humans evolutionary bred for good luck in the Ringworld books(?) (or more recently in that deadpool superhero movie)

[drummyfish]

Wow, I didn't watch the video but, did you get any FPS?

Yes, high FPS -- I am using Debian. AFAIK that is completely free, right? No non-free repositories enabled of course. That's even more free than GNU approved distros, because aside from SW they also enforce free data packages. But correct me if I'm wrong.

Strictly speaking you could probably find something on the blurry edge of HW and SW that's non-free in the PC and you could argue it's not open HW etc., but on the reasonable, practically possible level I think I can say it's all free.

the majority of the work going into the TAS is going to be luck manipulation

Can be -- depends on what it means of course -- e.g. RNG manipulation, or skill manipulation (e.g. the slowmo command I've presented above that gives you a 100% aim accuracy).

In a simple 2D platformer in an emulator the random generator can be manipulated relatively easily and you can usually see the ideal goal (get from left to right ASAP), but with native, non-linear 3D FPS games with a ton of variables you can't see the ideal path -- even just movement in 3D space with complex physics and a number of enemies makes this an equation whose ideal solution we won't be able to find, we have to rely on heuristics, which means skilled player intuition. You can do RNG manipulation to a degree of course, with the access to the source code and with debuggers, but I think the simple skill boost with time manipulation has a much bigger effect for the start. I've seen a GTA 3 TAS that seemed to be done in the same way, with time manipulation. RNG manipulation could fine-tune this at a later stage, but it would be a lot of work, planning, math etc.

I am now probably going to create a slowmo TAS of Xonotic, would like to see how much faster it is (and it's also hell of a fun to play like this!). Ideally for the next iteration savepoints, rewinds and other kinds of time manipulation tools would be great to be invented.

[onpon4]

I think the luck manipulation would be a similar complexity to Star Control II, which there is a TAS of:

http://tasvideos.org/1493S.html

AFAIK that is completely free, right? No non-free repositories enabled of course.

Yep, it is. And yep, it rejects the GNU FDL with the proprietary clauses it supports from the main repo.

The only reason the FSF doesn't endorse it is because they don't want their endorsed distros to "recommend" proprietary software by doing something like documenting how enabling the non-free repo works, or having an add-on repository for a Web browser that contains some proprietary add-ons. Even Pip is on the chopping block in Trisquel (or I think it's already been purged from the repo), and ScummVM has similarly been removed for such petty reasons.

[Lyberta]

Deleted.

[onpon4]

Intel graphics don't require proprietary firmware.

There isn't a libre game out today that requires anything an Intel integrated graphics controller isn't capable of.

[drummyfish]

TBH I didn't know which GPU I was actually using, I installed Debian on my older Desktop and it just worked like that, games running very smoothly (which wasn't the case in a previously tested Trisquel where they were tearing pretty badly).

I have both Intel and NVidia:

{l Code}: {l Select All Code}

lspci | grep VGA
00:02.0 VGA compatible controller: Intel Corporation 4th Generation Core Processor Family Integrated Graphics Controller (rev 06)
01:00.0 VGA compatible controller: NVIDIA Corporation GK104 [GeForce GTX 760] (rev a1)


And it seems Nvidia is used, with the noveau driver:

{l Code}: {l Select All Code}

glxinfo | egrep "OpenGL vendor|OpenGL renderer"
OpenGL vendor string: nouveau
OpenGL renderer string: Gallium 0.4 on NVE4


[onpon4]

You can't use both at once; the Nvidia card is in, so that's what's used. You'd have to physically pull the Nvidia card out to use the integrated graphics.

And your previous experience, as well as your experience with Trisquel, is most likely just the version of Linux you're using. The slow experience would be because your card wasn't supported by Nouveau yet, and the tearing going away would be because of improvements since the version of the kernel Trisquel is using.

[Lyberta]

Deleted.

[XGenGamer]

Some games I am interested in speedrunning are Mario and Luigi Superstar Saga and Pokemon Mystery Dungeon Explorers of Time/Darkness. What I am looking for in a speedgame is something that can be completed fast and has a active and moderately sized community.

[drummyfish]

You can't use both at once

Of course. On some laptops you can still switch between them for different programs, I used to use bumblebee for that, but it's a huge mess, not recommended.

This is an OT rant, possibly for another thread: I think all this graphics and GPU hype is doing no good. There should be at most one GPU in a typical computer -- graphics would be completely okay to stay at the level of PS1, with the focus on good art and efficient design (example: Crash Bandicoot), and most games should therefore run fine with just a software renderer. Proprietary gaming is setting completely opposite direction of course, but maybe in libre games we shouldn't try to compete, but instead lean more towards the opposite, the healthy way. I am currently working on a SW renderer I'd like to utilize for this purpose in my future games. Once it's in a presentable state, I'll create a thread and we can discuss this in more depth.

Some games I am interested in speedrunning are Mario and Luigi Superstar Saga and Pokemon Mystery Dungeon Explorers of Time/Darkness. What I am looking for in a speedgame is something that can be completed fast and has a active and moderately sized community.

Yeah, there is a bazillion games like that in the proprietary world, you will have no problem finding a community there. But maybe the real challenge, the one that deep inside you really want to pursue is helping start a new speedrunning community? The dark side is strong and tempting, you know... only the strong can resist!

[GunChleoc]

You can't use both at once; the Nvidia card is in, so that's what's used. You'd have to physically pull the Nvidia card out to use the integrated graphics.

Or call

{l Code}: {l Select All Code}

sudo prime-select intel

I haven't checked if that command is 100% libre though.

I am not using my NVidia at the moment because of driver issues, and I did not have to open up my laptop to remove it.

[onpon4]

I only could get stable FPS in Xonotic on ~max settings when I upgraded to Ryzen 7 2700X and Radeon RX 590.

That's a key point, though, isn't it? That's max settings; lower settings run fine.

This is an OT rant, possibly for another thread: I think all this graphics and GPU hype is doing no good. There should be at most one GPU in a typical computer -- graphics would be completely okay to stay at the level of PS1, with the focus on good art and efficient design (example: Crash Bandicoot), and most games should therefore run fine with just a software renderer. Proprietary gaming is setting completely opposite direction of course, but maybe in libre games we shouldn't try to compete, but instead lean more towards the opposite, the healthy way. I am currently working on a SW renderer I'd like to utilize for this purpose in my future games. Once it's in a presentable state, I'll create a thread and we can discuss this in more depth.

FWIW 3-D games have the peculiarity where things look more and more blurry the closer you are to them, so I can understand the trend. But yeah, ultra-high detail isn't really necessary.

An interesting thought is that if a game were to not use raster-based textures, it wouldn't have that particular problem. I don't know how well vector-based textures work, but if it did work well for a particular game, that would be really nice. Then when you got too close to an object it wouldn't suddenly look like a horrible blur, it just wouldn't gain any extra detail.

Of course, I've never worked on 3-D games before though, so I'm kind of talking about something out of my league.

[drummyfish]

An interesting thought is that if a game were to not use raster-based textures, it wouldn't have that particular problem. I don't know how well vector-based textures work, but if it did work well for a particular game, that would be really nice. Then when you got too close to an object it wouldn't suddenly look like a horrible blur, it just wouldn't gain any extra detail.

Of course, I've never worked on 3-D games before though, so I'm kind of talking about something out of my league.

Vector-based textures sound like a good idea and the kind of oldschool creative thinking (which went e.g. into the Doom engine), which I would like more people to have... instead of simply saying "let's make GPUs 100 times faster and expensive and simply use HD textures". Procedural textures or "smart stuff" like fractal-interpolation can be also experimented with to prevent blurrieness... of course it won't beat HD textures, but that's not the goal... HD hype is a learned thing pushed for by the market, but it is not what makes games or even movies enjoyable. At least I believe so.

To get back on topic: I'm now rendering my TAS and will post it soon hopefully. It's not a perfect run and it's not THAT much faster (mostly due to unskippable things), but I think this one will be much more fun to watch, the bullet time creates some really ridiculous moments.

[drummyfish]

Here is the Xonotic TAS -- as the description says, more of a proof of concept, first iteration, not nearly a close-to-perfect-run, would like to improve it, or inspire someone to improve it under 30 minutes. Still I think it's already fun to watch, but judge for yourself:

https://peertube.mastodon.host/videos/w ... e7a8141de4

Watch e.g. 15:05 -- it's a deathmatch where weapons change periodically, but it's won before even the first change happens.