EtaoinWu

If you have ever spent time on the Internet, you must have already known that people play video games, and you should have noticed that people try to play game fast. They are called speedrunners. If you spend your time similarly to how I spend mine, you could have watched a series of videos collectively titled Devs React to Speedruns by IGN.

The premise is exactly as its name suggests – a game developer, sometime multiple developers, watch someone speedrun their game for the first time, and realize the beautiful cinematism and level design they spent years painstakingly crafting being bypassed, skipped over, or sometimes destroyed. Some use physical collision glitches to make the player fly over half of the map, usually with commentary “how did we not find that??”; some parkour and walk on vertical cliffs, usually with commentary “you can do that??”. The devs’ reaction usually consists of a broth base of surprise, a tablespoon of frustration, and a dash of horror. The speedrunners are usually so good at what they do, that viewers may assume they skip a beautiful part of a perfectly good game just because they don’t like it or something.

Bennett Foddy has a different read. Developer of Getting Over It and Baby Steps, and simultaneously a philosophy PhD, he recognised that breaking a game is hard. With his smooth, almost painful-to-hear voice, he witted in his reaction,

The speedrunner takes that sculpture, and they look it over carefully, from top to bottom, from every angle, and deeply understand it, and they appreciate all the work that went into the design: all the strengths, all the weak points. And then, having understood it perfectly, they break it over their knee.

Being a speedrunner myself (of the wonderful 3D puzzle game Superliminal and more), I appreciate the appreciation of speedruns by Foddy. People seem to have divided ideas about what speedrunning really is; in this blog, I attempt to explore some possibilities.

Speedrunning is misreading

The frustration of the game devs is not unique to the media of games.

The book Fahrenheit 451 has been long understood by most as a clever attack on content censorship, but its author Ray Bradbury has famously got into into an argument with a student when presenting at a university, when Bradbury insisted he intended the book to be an attack on mass media, specifically television. The argument was so fierce, according to the legends, that he literally stormed out of the lecture hall. His frustration must be homogeneous to the game devs – the feeling that one’s intellectual work was ignored and its beauty skipped. Neil Gaiman, on his introduction to the 60th anniversary edition of the book, wrote:

If someone tells you what a story is about, they are probably right. If they tell you that that is all the story is about, they are very definitely wrong.

The phenomenon here is widely popularised by Roland Barthes as the “Death of the Author” in literary theory. Barthes argued that, when a text is materialised from the creator’s mind, it loses its connection with the intention of the creator, and its meaning and significance lie solely in the text verbatim itself. The potential, and very often reality if the work is popular, of misreading frequently infuriates authors who realize their works hijacked through its own text. Similar to our example of the censorship reading of Fahrenheit, J.R.R. Tolkien explicitly wrote that he disagrees with the atomic reading of the Lord of the Rings, in which 20th century readers understand the One Ring as the one bomb of geopolitical mass destruction. He “cordially dislike[s] allegory in all its manifestations”, and his story “is not an allegory of Atomic power, but of Power”. Author and activist Upton Sinclair, of The Jungle fame, said he “aimed at the public’s heart” by trying to bring to light the inhumane working conditions of meat industry workers, but “by accident I hit the stomach” by causing a public uproar in the sanitation of meat.

The similarity between speedrunning and misreading even extends slightly to etymology – a text could only share ideas and invoke emotions through the reading of readers, just like a game must bring its experience to players through the action of running the game. An author designs their wording, sentences, narrative, and characters, for ideas and emotions to get to readers; it’s just like a game developer designs the levels, graphics, UI and animation to strike similar feelings in players.

Therefore, reading and playing in unintended ways unite through the lens of literary theory. On this front, speedrunning suffers a little bit less infamy because the word doesn’t imply being a wrong way of playing, unlike misreading; but certain viewers of speedrunning certainly thinks so.

There is, however, a significant mechanical difference between a text and a game. Texts are inherently fuzzy from the most micro, word level, making them unavoidably subjective, whereas a video game as a program can be interpreted as a strict, computable formal system. A game of fixed source code on a computer can be modelled as an abstract function \(f : \mathrm{Input}\times\mathrm{State}\to \mathrm{State}'\), and hypothetically every play interacts with the game in a effective way. Therefore, excluding marginal cases (like console gamers using filthy CD disks to get a faster time), unintended play in games is an even stronger notion than unintended reading of texts, because exploiting a formal system is harder than an inherently ambiguous one. A specific input to the function \(f\) that achieves a certain end-state is, within this system, no longer an opinion but a fact.

This means…

Speedrunning is a mathematics academia

Mathematical formalism, pioneered by David Hilbert in the early 20th century, is the philosophical idea that mathematics is essentially a game played with organizing glyphs on a paper according to strictly defined rules; the initial glyph sequences you have available are axioms, the manipulation rules are logical derivations, and the results are propositions, lemmas and theorems. Mathematical formalism is the epistemological telos of formalizing mathematics, abstracting away the potential link to reality. In this aspect, doing mathematics is no different to playing a video game, and the mathematical system is just another function of a similar signature, \(f : \mathrm{Derivation} \times \mathrm{Proposition} \to \mathrm{Proposition}'\). The complexity of a game engine and an axiom system can both be encoded as the function \(f\), and watching a playthrough is isomorphic to reading a maths paper.

A special type of speedrunning, known as Tool Assisted Speedrunning (TAS), fit our math metaphor even better. Tool-Assisted Speedruns uses emulators, savestates, and frame-by-frame inputs to construct a hypothetical optimal way of play, removing the human execution element. TASes abstracts away the fact that the game input must be performed (or performable) by a human, and aims to find a specific input sequence to achieve a goal, for example finishing Celeste as fast as possible, just like a mathematician aims to find a derivation sequence to achieve a certain theorem, i.e., a proof. In fact, the analogy goes both ways, as mathematics has in fact been gamified recently in the Natural Number Game.

This only explains a third of the three-word phrase in this section’s title. The central noun is “academia.” No singular person can grasp the entirety of modern mathematics, so an entire community is formed through collaboration, and later researchers can base their work on previous papers; speedrunning is also a community effort, where strategies (“strats”) and tricks are passed down through YouTube videos and Discord clips. Pure mathematicians do the theory, while applied mathematician use them; smart TASers and out-of-the-box-thinking glitch hunters design the exploits and routes, while determined and finessed speedrunners do the grinding, the muscle memory, and the performance under real-time pressure.

They even share the naming convention: Just like mathematical results are often named after their discoverers, like Bézout’s Lemma in number theory after Étienne Bézout, in speedrunning we have Boba Skip after boba_witch, and Demo Dashing after DemoJameson. Speedrunners also have the tradition of calling them discoveries, not inventions, just like mathematicians.

The last word in the section title is the indefinite article “a”. Just like different mathematical universes can be constructed with different axioms (parallel postulate, law of excluded middle, axiom of choice being the most famous examples), every video game is its own distinct system, and requires its own sub-community. Every runner can have their favour and skill set, making them affined to different sub-communities; but just like mathematicians, they can often span multiple sub-fields, because skills often transfer.

Speedrunning is emergence

Another source of anger in devs who react to speedruns is the regret of not finding and patching a glitch. Such glitches are unavoidable simply because a game is too complex. When a formal system is pushed to its limits, the interaction of simple micro-rules often emerge into unintended macro-behaviours. Very often, speedrun tricks are by-products of completely intended game mechanics. Game developers program in physics rules, state machines, and hitboxes, but even when each subsystem is perfect, unforeseen consequences can appear at their intersection.

Take the movement physics in Source engine games like Counter-Strike: Source, which gave birth to the bunny hopping and surfing. The developers intentionally limits the player’s velocity by applying friction when a player is on the ground, but allowed players to accelerate by holding a directional key (left or right) while moving their crosshair in the same direction. Players discovered that by doing smooth and calculated sideway strafing in the air can accelerate their model rapidly, and if the player immediately jumps after landing, the ground friction doesn’t have time to kick in, thus bypassing the velocity limit. Instead of a disrespect to developers like “oh haha stupid implementation”, the discovery birthed some entirely new movement-based game modes and a great enthusiastic community.

We see a similar kind of emergent behaviour in Minecraft. The game has the following three mechanics:

1. Taking damage gives the player a slight nudge to simulate the knock-back effect;
2. Players take periodic damage when on fire;
3. A standard jump has a fixed vertical velocity that makes you up 1.25 blocks, no matter how long you press the jump key.

These three mechanics are perfectly intended and each individually makes sense. However, as a combination, if a player sets themselves on fire and time a jump perfectly, they can get the knock-back at the optimal point in the jump, thus clearing a 1.5 block tall obstacle.

In both examples, the game is not broken, and the engine is running its designed logic flawlessly; the emergence is the realization that the game engine is agnostic to the intention or the physical reality, and operate on the only truth of its literal ruleset.

Speedrunning is jurisprudence

Emergence creates ambiguity. If combining three non-glitches result in a physically unrealistic movement, is the result a glitch? This question is particularly important to answer, as many games’ speedrunning scene has a separate leaderboard for “glitchless” speedruns, where players compete to finish the game as fast as possible without using game-breaking glitches.

Unfortunately, it seems there is no computer-scientific or mathematical answer to this question. Different communities have different opinions on what is a glitch and what is a clever use of intended game mechanics. In this way, “glitchless” categories are subjective, community-defined rule sets.

To understand this, we have to look at how these communities handle edge cases. A bug causing the player to clip through a wall is certainly a glitch; Flying by standing on an item you are currently holding violates Newton’s laws and certainly breaks physics. But if three perfectly normal intended mechanics are combined, like the damage-tick jump in Minecraft, and skips a part of the story, is this a glitch? What about using the pause menu to cancel horizontal velocity in Grand Theft Auto 5? What about logging out and re-logging in to get invincibility (e.g. in Minecraft)? Spawn invincibility is a perfectly intended, good-faith feature that aims to protect the player, but is it an abuse if you do it every 3 seconds to refresh it? If the category bans “glitches”, does it ban logging out (or pause menus) entirely? The true answer always lies in the middle and has nuances.

This video by EZScape explores this topic nicely, which I will steal some examples from. Mirror’s Edge has a “glitchless” category that freely allows unrealistic physics exploits like wall-boosting and fall-damage cancelling, and they are important components of the run. The Legend of Zelda: Ocarina of Time allows the use of a bomb knockback to boost the player across a gap, which is similar to our Minecraft example above; the game’s community also allows the “Power Crouch Stab”, a bug move that is allowed simply because it’s a simple button input and easy to trigger accidentally.

In this way, the speedrun community and its moderation members form a judicial body. They have to debate on forums and Discord servers, like real-world lawmakers, to draw the line somewhere.

Any% runs, i.e. finishing the game with whatever means necessary, is a reflection of constitutional textualism, which asks: What does the codified text, i.e. the law and the game, allow? This treats the game as purely a formal system (as described above). However, even this seemingly formal and rigorous viewpoint breaks when players discover Arbitrary Code Execution (ACE). In certain games like Super Mario World and the previously mentioned Ocarina of Time, speedrunners realised that the source code contains bugs that can be exploited from not a gamer, but a computer security expert’s perspective, that allows overwriting the game’s memory. This gives the player infinite ability in modifying how the game is run, for example turning the game into Snake or Flappy Bird (video here); obviously, using it to complete the game is as easy as jumping to the end credits. Even the most liberal speedrun critic will probably think that this is too much; we are not beating a game anymore, but hacking a software.

On the other side, we have originalism, which asks: What was the developers’ intention? But as we have established with the Death of the Developer, speedrunning itself is deeply unintended. Ernő Rubik’s original intent for the Rubik’s cube – a visual tool for 3D geometry – has zero effect on how the World Cube Association govern nowadays. Apart from a few examples, the intention of a game developers is not for the players to speedrun at all. Using the original intention to regulate an inherently unintended way of play is less than useful.

Further, even when the original intention is interpretable, the context of the origin may have already been lost. Take gun rights in the US as an example; the second amendment intended to grant rights for people to bear arms, but the flintlock muskets of the 18th century is no match for a 21st century semi rifle. Whether or not the centuries old amendment cover the firearms of today is one of the main fronts of debate in gun rights now. Similarly, if we consider a Super Nintendo game as a piece of deterministic software (encoded as a function \(f\)), do we allow the player to press the left and right buttons at the same time? The original hardware has a solid D-pad that doesn’t allow this kind of input, but porting them to a modern arcade box or a simulator on a PC will certainly break this constraint. The community must decide on whether the physical limitations of the 90s plastic are binding on a 2026 leaderboard.

The principles of the communities, as EZScape noticed in his video, seem to centre around several values: playability, randomness, and skill ceiling. The speedrunner communities usually decide to rule out tricks that make the game too easy, too hard, or too random. They aim to make the category a beginner-friendly leaderboard but with certain competitive features that makes it enjoyable to play and reflects players’ skills. This means …

Speedrunning is sports

The reason why communities have to make strict rules on what is allowed is because speedrunning is a competition; the goal of the competition is literally in the first half of the word itself. It creates a PVP (player-vs-player) meta-game out of a PVE (player-vs-environment) base-game.

This is the category ruling needs to be thoroughly debated, as discussed in the last section; a deeply optimised sport needs a fair playing field. Any hole in the ruling may and will be exploited to their maximum extent. Just look at motor sports: when the rule on banning ground effect only specifies the static ride height, the 1981 Brabham created a mechanism that only lowers the car when it is at speed and the downforce kicks in, essentially bypassing that rule.

Speedrunning is a competitive sport, not in the way soccer is, but more similar to track and field; players are not playing with, but still competing with, each other. This is sometimes known as a “closed skills” sport. If the TASing part is pure maths or theoretical physics, executing the run is Olympic hurdling, marathon, or gymnastics. The physical execution requires top runners to invest thousands of hours to train their hand-eye coordination and muscle memory, and the most extreme speedruns requires immense physiological and psychological endurance to pull off.

And the grind is truly immense. For games like Trackmania, an arcade racing game, players will drive the exact thirty-second track tens of thousands of times, trying to meticulously achieve certain racing lines to shave off sometimes 0.01 seconds on a record. If you are familiar with Wirtual’s documentaries, it will not unfamiliar to you to hear a record being broken “after three months of nonstop grinding” of a player. The grind is even more significant in retro games, where the playing field is already heavily optimised. A Super Mario 64 runner aiming for 120 stars essentially runs a marathon, but each minute of the two hour journey requires sprint-level reflexes; a single dropped frame at the wrong time could kill a world-record pace. In the classical NES Tetris scene, players have to play with reflex and finesse that’s unbelievably fast even from several years ago, and to do this, they have to physically change how they grip their controller from standard tapping to “rolling” (alternating between multiple fingers on the back of the controller), bypassing the biological limits of human fingers. These optimizations are the result of generations of speedrunners contributing to the same competition.

The competitiveness of speedrunning is also why massive spectator-filled events are happening in the speedrunning scene. Events like ESA (European Speedrunner Assembly) in Europe and GDQ (Games Done Quick) in the US welcome thousands of visitors, watching, reacting and cheering for speedrunners show their amazing skills. Even when they are not familiar with the games, like most sports fans not really familiar with top-level sports plays, they still enjoy the event in the same way. The cheering and ovation of GDQ watchers when a runner pulls off a frame-perfect pixel-perfect trick is no different than a stadium watching a home run, or a three-pointer from half court. It must be an unforgettable experience, for the players and audience alike.

Therefore …

Speedrunning is fun

This is the ultimate reason why people speedrun; Speedrunning is the gamification of playing a game.

This is probably the deepest mis-intention of speedrunning. Video game developers have the intention to create a specific, carefully designed type of fun. The player’s intention, however, is just to be entertained. The two incentives do not necessarily align.

You can see this in casual, highly social contexts. In Counter-Strike 2, a team competitive first-person shooter, on the map de_dust2, there is an interactable football (soccer ball if you are American). When a match is hopelessly lost and the game stops being fun, sometimes a team will propose in text chat to play a makeshift game of football (soccer) in the “Mid” part of the map, using the map’s doors as goalposts.

There are no glitches or bugs involved here. It relies entirely on the two teams’ good faith, on the true reason people play games: to have fun. However, this deeply violates the entire idea of the game, which is supposed to be a life-or-death military fight between two opposing forces. This proves that humans will come up with rules, and sometimes ignore them, just to entertain themselves.

This appears in multiple base-games, and is most visible in games when explicitly designed to not have rules. In Minecraft, a voxel-based survival game, people create parkour maps, music-synchronized minecart tracks, and automatic electromechanical gates. The Legend of Zelda: Tears of the Kingdom, an action-adventure game that gives players building tools intended to construct bridges and carts to save the kingdom, sees an amazing community within 48 hours of its release, building multi-stage orbital strike laser canons.

Wikipedia has a marginally related article on topic; players may find creative solutions, play with certain constraints (going sideways only, playing with the worst character, or playing an assassination game without killing anyone, etc.), or play for a second objective. Time is universal and a natural choice for such an objective. When a player really appreciates a game, they spend more time to try to re-play it, and thus naturally getting better at it, often getting faster and faster playthrough. When they are so familiar with it that they have explored all maps, and familiarized themselves with every decision they need to make, time is the natural objective.

Thus, this is also the biggest misunderstanding from some viewers of speedruns, including some developers in the IGN series: Speedrunners are not skipping sections because they dislike them, but because they enjoy the game so much that they want to make more entertainment from it even after hundreds of hours in it. It is because a deep appreciation of the game; after all, why would one spend an entire week practicing a specific trick in a game if they hate it?

So, yeah. We are humans and humans entertain themselves whatever the obstacle is, even in a rebellious way. The developers can build a function \(f\) with an intended input \(x\), but however they hint at the intended solution, they can never entirely dictate what the end user decides to do with it, or what fun we may have from it. What’s important is that we have fun.

And let us have fun.