What is a Speedrun? Categories, Glitches and World Records

Summary: A speedrun is an attempt to finish a game as fast as possible. It is a competitive discipline with its own rules, categories and verification, not just playing quickly. Runs are grouped so like is compared with like, glitches are usually allowed in the headline category, and records are checked by volunteers against submitted video before they stand.

• A speedrun is a timed completion attempt, judged against a defined category.
• Categories exist because rules change what the challenge even is: Any Percent, 100 Percent, Glitchless and many more.
• Glitches are generally legal unless a category bans them. Exploiting the game is part of the craft.
• Records are community-verified from video evidence, and rejected submissions are common.

Speedrunning turns finished games into new ones. A title people played for forty hours gets dismantled until it can be completed in twenty minutes, and the interesting part is how the community agrees on what counts.

Where Speedrunning Came From

Players have raced games since arcades, but the organised form grew out of 1990s PC shooters, where recorded demo files could be shared and verified by anyone with the game. That is the important detail: the hobby needed a way to prove a time, and it grew wherever proof was possible.

Cheap video capture and streaming turned it from a niche into a spectator activity, and charity marathons pushed it further into public view. The infrastructure changed the hobby more than the games did.

How Runs Are Timed

Two clocks matter. Real time is wall-clock from a defined start to a defined end, which is simple and unforgiving. In-game time uses the game's own timer, excluding loading screens, which makes comparisons fairer across different hardware.

The distinction is not pedantic. A fast machine can save minutes of loading, so leaderboards for the same game often keep both timings. It is one reason hardware discussions matter in the community, from custom loading setups to handhelds like the Steam Deck.

Where the timer starts and stops is defined per game and argued over in detail, because a second of ambiguity invalidates comparisons.

The Main Categories

Categories are the rulebook, and the same game supports many:

• Any Percent: reach the ending by any means. Usually the fastest and most glitch-heavy category.
• 100 Percent: complete everything the game counts as completion, which is defined per title.
• Glitchless: finish without exploits, which often produces the run closest to intended play.
• Low Percent: finish with as few upgrades or items as possible, which is usually harder than Any Percent.
• Individual Level: a single stage rather than the whole game, which is the easiest way to start.

Categories multiply because they are how the community keeps a game interesting. Once one route is solved, a stricter ruleset opens a fresh problem.

Glitches, Skips and Why They Count

Newcomers often assume exploiting bugs is cheating. In most categories it is the opposite: finding and executing a glitch reliably is the skill being demonstrated. Clipping through a wall on demand can take hundreds of attempts to learn and remains difficult under pressure.

The genuine line is external assistance. Modifying the game, using scripts or macros, or splicing footage from separate attempts is cheating everywhere. Using what the game does on its own, however unintended, is not.

How Records Get Verified

A submitted run is video evidence reviewed by volunteer moderators against the category rules. They check the timing points, look for edits or splices, confirm the settings and version used, and sometimes ask for additional footage such as the loading of a save file.

Rejections happen for honest reasons more often than dishonest ones: a wrong game version, a missing timer, an unclear cut. The community has also caught several prominent fabricated runs through frame analysis, which is why the evidence standard keeps rising.

Tool-Assisted Runs Are a Separate Thing

A tool-assisted speedrun uses emulator features such as slowdown, frame-by-frame input and rewinding to construct a theoretically perfect run. These are never compared against human runs and are not presented as such. They exist to show the game's absolute limit, and they frequently reveal techniques that humans then learn to perform live.

Trying One Yourself

The barrier is lower than it looks. Pick a game you already know well, choose an Individual Level or Any Percent category, read the rules for it before starting, and use a free split timer. Your first serious attempt will be slow and that is the point: you are competing against your own previous time long before anyone else's.

Short games with tight controls are the usual starting point, and there are plenty among the free games on Steam if you want to try the hobby before spending anything on it.

When you do want a specific title to run, browse gaming gift cards and top up a store balance rather than putting a card on file.