Fun to see Cosmo on Hacker News. He's been getting a fair amount of attention recently, including an article on Yahoo!. If you're interested in seeing him exploit more video game glitches in the name of speedrunning, you can watch him on Twitch.tv at <http://twitch.tv/cosmowright>. He is one of the (and occasionally the single) best Wind Waker runners in the world. Right now he's concentrating on The Wind Waker HD.
Cosmo is one of the best streamers to watch for those who question the legitimacy of watching gaming as a form of entertainment. He is an entertaining and interactive host, and you can see very tangible effects of his countless hours of efforts learning to break these games.
For those who haven't seen it, I highly recommend one of the "original" speedruns - Quake Done Quick (on Nightmare, naturally). It's less about glitches and more about playing extremely well. If you have ever played a fast-paced FPS, you'll appreciate it.
Quake Done Quick is an ultra classic, and it's awesome it's gained enough fame to get its own Wikipedia page . It was amazing way back in the day, but it's still no less amazing.
My little claim to fame is I used to play QuakeWorld quite regularly with one of the founders of QDQ. Before it was called Quake Done Quick, it was called "Nightmare Speed Demos" (http://web.archive.org/web/19970719050959/http://www.planetq...). Radix (who at the time, went by "Lord of Radix") was a member of the "Deadly Venom Clan", and was one of the last keyboard-only holdouts in Quake, and despite using the keyboard for everything in Quake, was still a force to be reckoned with in deathmatch.
Anyway, I always smile when I see QDQ mentioned on the internet. Every few years, I dust it off an watch the whole thing and amaze myself in how precise all those actions are. Quake was such a fantastic game mechanically, and I wax nostalgic over the long-lost sense of community it bore, compared with the ridiculously disrespectful nature of internet nerd assholes these days (It probably doesn't help that I mostly play League of Legends these days).
In most speedruns, player runs away like a sissy little girl. In this speedrun, players aim for 100% kills and 100% secrets. Secrets actually make it more interesting to watch, as they often store the Quad Damage powerup.
If we're going all the way, "little" may be offensive to people with a growth disorder, and "child" is age discrimination, the whole combination implying that only children can be small and scared.
tl;dr, stop being so politically correct everyone, sheesh.
Because it's important for us to root out sexism where we find it, for the same reasons we pull up any obnoxious weed. The fact it wasn't intentional makes it all the more obnoxious.
Yes. Sexism is terrible and must be eridacated. I thank you for your hard work making the world a better place (for yourself), and suggest the following to make sexists even easier to root out:
Make all sexists sew a yellow male-sign onto their clothing.
I think he technically meant to say 'sissy little boy.' The word sissy is not necessarily 'antagonistic,' at least according to its accepted definition of "weak and fearful."
Funny, I just watched that run again last night. My favorite stages are the ones where enemies are killed after the stage is finished. That technique shaves a few seconds off the overall run.
To be honest, I think I prefer learning about how people have discovered and exploited the underlying mechanics of a game. I really enjoyed reading the Pac Man Dossier, for example:
There are a lot more than that one, too! QDQ is the original, but Quake done Quicker and the Rabbit Run are also required viewing. And then you move on to things like Super Metroid... It was this stuff that got me into the Speedruns (and TAS) scene. Your two main resources for this are:
for "tool-assisted" speedruns or time attacks, meaning using emulators and often doing truly insane frame-by-frame optimization and memory-state-manipulation.
One of my favourites; not technically a speed run, but a festival of glitches being used to corrupt memory in a very specific way, using only the regular gameboy inputs :)
Even though it's not the fastest, I still love this version "Quake Done Quicker", in which the recordings were edited to show the player in third person and includes "snappy" dialogue and slow motion sections.
It's more about having memorized every map, including the locations of all enemies on the map. I wonder how well someone could do in a game where enemies were so predictably placed.
Basically there's a speedrunner who is technically adept at the game, is a good communicator, and explains as he plays a complex exploit that facilitates skipping a large part of the game
I was watching that video and expecting an explanation along the lines of "If I tap the c-pad, then b really fast at just the right time" and being met with an explanation that shows an intricate working knowledge of the game....it was impressive to say the least.
Wow. That guy is a hacker. Maybe he is only a hacker of this one game, I don't know. But, this is a video that deserves to be linked by HackerNews.
"Realizing this ROM hack is able to control the CPU with only the player’s inventory, [bortreb] wanted to see how far he could push this hack. He ended up writing a bootstrapping program by depositing and discarding items from the in-game PC, and was then able to reprogram the Game Boy with a number of button presses on the D-pad, select, start, A and B buttons."
Not to diminish Cosmo's skills or his excellent explanations, but the tricks used in these speed runs were discovered and improved upon over many years by many different members of the speed running community who all deserve some credit. [1] has a list of techniques and usually notes who originally discovered them. There are some speed runners who are very good at discovering these kinds of techniques but they are not always the ones who are the fastest at running the game.
While the exploit itself was very impressive, I was also very impressed by how clearly he explained it, and how he was able to do it without stumbling, and generally was able to keep playing while explaining. Great communication skills.
The video is a clip from the Awesome Games Done Quick 2013 charity marathon[0]. It's fun to watch, and Cosmo in particular typically goes into detail on how the weird tricks and hacks he performs work.
I love looking at glitches in older games that had so little memory protection. You can learn a good deal about how a computer operates by dissecting them.
This run doesn't even use the latest tech (or he might intentionally be playing a slower-but-safer route to facilitate commentary); I think he shaved another three minutes off by --
* tightening up the run to the village (by supersliding sooner)
* not needing to die on Gohma (you can glitch through the dungeon door that closes)
* skipping the tower collapse entirely (by falling through the map, which he discusses briefly as "impossible for a human to do" in the parent's video).
The current WR doesn't use the TAS method he describes in the video, it is similar through, but only requires getting hit by 1 rock, not 2. And doesn't require the slingshot. The "TAS-only" trick skips loading the inside of the tower and basically ends up at the bottom or something. The trick in the WR gets hit by the rock and you end up in the tower but skip to the end of the inside part.
I think you're confusing the glitches a bit (or I am). Long-winded explanations of both, assuming I understand them correctly:
The TAS version does a trick in the Deku Tree that results in the one set of bars in the tower escape being up. There's a bitmask in memory that dungeons can use to store flags about whatever they want. One of those flags in the Deku Tree dungeon is whether the ladder in the slingshot room has been shot down. That same flag in the dungeon escape indicates that Zelda has already raised a particular set of bars at one of the tower entrances. Because you warp straight from the tree to the tower, it keeps the flags set, so if the ladder was shot down in the tree the one set of bars is already up in the tower and you don't need to wait for Zelda to open them. This saves some time because she's slow, but in theory you'd still need to run all the way to the door.
Luckily, there's a second glitch that involves falling off the edge of the tower. Normally the game detects when you fall too far and warps you back to the top. If you hit a boulder in mid-air, it resets the fall counter, and you can manage to fall off one bridge and land on the bridge below. The TAS does this twice to fall down to the bridge that has the open gate, so Link can then run right through the entrance without waiting for Zelda (she catches up when you enter a new area). This is the part that Cosmo was describing as impossible for a human, because the boulders are random so you can't really time it such that you hit one in mid-air at the right spot to reset the counter.
The void warp (the new trick used in the WW) is kind of related in that it involves falling off the tower, but that's about it. The entire setup is such that the game detects that you've fallen off the bridge at the exact moment that you hit the loading zone for the inside of the tower. The game tries to reset your position back to the bridge you fell off of, and tries to put you inside the tower, so you end up in the tower but outside the normal rooms, so you just fall all the way to ground level (which is where you wanted to be anyway) and walk to the next loading zone
I love speedruns. There's a vibrant community of speedrunners and fans out there. With fast Internet connections, they can now stream their attempts live.
Just a heads-up to anybody interested -- there are generally two types of speedruns, tool-assisted speedrun (TAS) and unassisted. Searching for "$GAME TAS" (eg: Super Mario World TAS) usually results in very impressive videos so they're usually easier to find.
There are a variety of other terms as well, for example sequence breaking, no-glitch, 100% completion, and so forth. Understandably there are numerous videos on Youtube for pretty much every console game that can be played on an emulator.
By far my favorite kind of speedrun to watch is a non-TAS with commentary. Often, these runs are done by teams. There's typically one player who is good at executing, and then you might have someone responsible for planning the route through the game, and another person responsible for researching glitches and how they can be used most efficiently, etc.
Hearing these people discuss their craft, and hearing the dynamics between their interactions (personal and practical) is fascinating. The best example I know of this is the ingx24 speedrun of Majora's Mask: https://www.youtube.com/watch?v=85k0fo1qnyo
Humans are amazing animals. The fact that this level and coordination is possible even when the stakes are, ultimately, pretty pointless, is a testament to that.
Also, see Speed Demos Archive (http://speeddemosarchive.com/), one of the best places to watch existing runs on a variety of games. The forums are still a good community, although as vinkelhake mentions most people use twitch to stream practices and runs.
Yeah, and one of the biggest value-adds of SDA is that the runs you see up there are verified by other runners. It takes a long time for a run to get up there, but once it is, you know it's the real deal.
There's also a cool, recent exploit in Super Mario World speed runs that uses similar RAM corruption techniques to "beat" the game in less than 3 minutes. (more details here: http://minimaxir.com/2013/03/127-yoshis-in-slot-6/ )
The technical details of how this series of exploits were identified, much less applied, is staggeringly complex. I love speed and/or glitch runs and I can't imagine this one being dethroned anytime soon.
Actually, it already was. The fastest Super Mario World speedrun is now 1 minute and 40 seconds. For more details see http://tasvideos.org/2380M.html. It's even more complex.
"It probably helps that I’m guessing this game had to be rushed out the door and perhaps had some inexperienced programmers working on it, which led to a more fascinating and strange world for speed-runners to explore."
Every piece of software has bugs. But, the kinds of things they exploit in OOT you wouldn't find in %99.999 of normal playthroughs.
Yep, there are a ton of huge / game-breaking bugs. But it did take quite a while (10+ years) and a lot of people to find many of them. The fact that people really love to speed run the game certainly was a huge factor in finding these subtle bugs, not to mention technology like emulators really accelerated the process, especially when it came to finding, for example, the correct frame to leave Gohma's room for the wrong-warp to Ganon's castle (it's a frame perfect trick). I would imagine that most games made at the time, if subjected to the same level of scrutiny, would display near or the same level of brokenness. Donkey Kong 64 certainly comes to mind. I don't think this is indicative of inexperienced programmers.
And then we have to think of the system limitations at the time and the fact that the N64 was a pretty new system with a new way of displaying graphics. It was a pretty big paradigm shift and yes, for the first 3D Zelda game it, like all games, could have benefitted from more time.
A lot of the bugs with the game have to do with messing with memory and exploiting pointer arithmetic. Using pointer manipulation to handle cutscene / area changes as well as items on the C-buttons obviously isn't ideal but sure as hell saved space compared to alternative solutions. There is very little error checking (for buffer overflows and 'impossible' situations such as having a deku stick on 'B') for what I would imagine to be performance concerns as well as simply good time management.
To your point, absolutely. Very few people for the first few years encountered any such bugs. The first real speed run of the game came in 2003, 5 years after its release. Even that run featured very few bugs; it was mostly route optimization and practice. The latest wrong warp glitch which enables you to jump from Deku Tree to Ganon's Castle was found in 2012. That's a full 14 years after the game's release.
If anything, this speaks more to the dedication of speed runners and glitch finders that the incompetency of OoT programmers.
I don't even think that it was inexperienced programmers. Ocarina of Time was the biggest game world ever created at the time, and it was released on a 32MB cartridge.
32MB. Every single bit of that game is exactly where it needed to be.
I came here to say this, basically. The glitches exploited here are perfectly acceptable , and it would be really weird to do the QA required to get rid of them.
If you are interested in seeing more of the best speedrunners, Awesome Games Done Quick is happening in January. It is a week-long 24/7 marathon of speedrunning to raise money for charity: http://marathon.speeddemosarchive.com/upcoming
This is an interesting subculture of gaming: people trying to dissect a game by all means possible. For Shadow of the Colossus, there is a community dedicated to finding every bit of the game world, even the unfinished parts that made it into the final versions:
I just recently got into the speedrunning scene, and it's interesting how the community categorizes speedruns based on version differences and platform. Cosmo wrote a great blog post specifically about OOT: http://blog.cosmowright.com/?p=33
Predictably, this speed run is now quite outdated, even though it's less than a year later. There have been several major discoveries that have now lowered the world record to just over 19 minutes.
A more complete explanation of the glitch occurs earlier in the video, but he explains the basics of it after he performs the glitch (unfortunately dying in the process). Apparently three different exploiters found three exploits that, when combined, enabled this glitch. Quite fascinating.
Speedruns are amazingly interesting. When I tell people about watching them, their reaction is usually something along the lines of "why do people waste their lives on that?"
But there's so much interesting going on there. The parallels to optimization in programming are striking.
Sorry to be that guy, but isn't one of the cornerstones of optimization in programming that you should only do it when it actually gives a meaningful benefit? E.g. don't spend days or weeks to make a part of the program that gets called very rarely run 0.1% faster, when you could instead spend that time on solving a new problem, or making an actually useful optimization, or doing something else entirely.
Of course, it's silly to bicker about what is "useful" when it comes to things people do to relax or entertain themselves; and usually fruitless when it comes to things they do as escapism or getting a feeling of achievement to compensate for areas where they are stuck or blocked off. I do think the psychology of it is interesting though, and my personal guess is that most if not all people have the innate desire to progress and achieve things, but we often can't for some reason or other. Why does a caged animal endlessly walk in circles? Because it's better than doing nothing. And before you think I'm just being snarky, think again, please; I did and do my fair share of such things myself, I just also can't help but notice that the smaller and more powerless people become, the more we celebrate rather pointless things - TV and the interwebs are full of that. And even if I'm wrong about the motivations of everybody else on the planet, I do know myself well enough, and although I might be projecting, it's at worst a honest mistake.
1.) "premature optimization", the fact that optimization for the sake of optimization isn't optimization, but rather counterproductive, if not obsession or procrastination.
2.) my attempt to answer the question "why do people do this?", which is something I ponder a lot, in relation to gaming and other things as well. Imagine someone cooking the same meal over and over again, to optimize the amount of time they take for it, and the amount of ingredients they use, but never actually eating what they cooked.
Shouldn't a true speedrun be about going as fast as possible through the intended game, without exploiting obvious bugs? Finding those bugs is really cool, but I wouldn't call the result a speedrun.
If you liked the video in the article, you might also enjoy his commentary on a full Wind Waker run (nearly 5 hours): http://www.youtube.com/watch?v=0u3djy6Ednc