The Super Mario 64 world record board reads like a menu of different sports: the 70 Star record sits past 47 minutes, the 16 Star record under 15 minutes, and the 1 Key (0.5 Star) route ends in minutes. Same game, same cartridge, radically different skills. That is because speedrunning does not rank times on a single ladder — it ranks them inside categories, each with its own definition of what counts as finishing and what counts as cheating.
Understanding those rules is the difference between watching a run and reading it. Here is how the category system works, why leaderboard boundaries are drawn where they are, and which category suits a first-time runner.
What does the category name actually constrain?
Every category answers two questions: what must you complete, and what may you exploit? The common shapes:
- Any% — reach the end credit screen by any means, skipping as much of the game as routing allows. The name is borrowed from the completion percentage display in Metroid-era games; the point is that completion percentage is irrelevant. This is usually a game's flagship category and its most broken one.
- 100% — collect every required element: all stars, all temples, all Chaos Emeralds, per the game's own completion counter. Routing length balloons, and skips that would jump the counter are illegal.
- Glitchless — the credit screen must be reached without major unintended movement, per the community's written ban list. The term is a misnomer everywhere; 'minor' glitches like blind jumps are legal, while sequence breaks that leave the map are not. Every glitchless board is enforced by a rules document, not by physics.
- Low% — the mirror of 100%: finish with the smallest completion stat possible, usually the hardest category in any game that has it.
- Individual level (IL) — time a single stage or boss rather than the whole game, the sprint event of the calendar.
Leaderboard services like Speedrun.com host the boards, but the rules themselves are community-written documents maintained by moderator teams per game. That governance structure — volunteers publishing category rules, verifying video evidence, ruling on edge cases — is the actual institution of speedrunning, older and more stable than any event around it.
Why does one skip split a community?
Category boundaries exist because communities decide what the run is 'about.' When a new skip is discovered, it does not merely improve times — it can invalidate every route on the board. The pattern repeats in every game's history: someone finds a sequence break, the any% meta rewrites itself within weeks, and the community either absorbs the new route or fences it off into a new category. The discovery of the 'Checksum' skip tier in Mario 64's routing culture and the wallpaper clip in The Legend of Zelda: Ocarina of Time are textbook cases: each redrew its game's map of possible runs, and each forced moderators to decide whether the game was being finished 'as designed' or merely finished.
This is also why timing rules matter. Most communities time real-time (RTA), start to credits as the video shows; some, particularly older JRPG communities, time load-screen-adjusted or in-game clock. Comparing a runner's time across games, or across timing conventions, is a category error in itself.
Related stories: From Itch.io Demo to a Million Copies: How Small Games Actually Get Found · How a Weeklong Speedrun Marathon Raised Over $50 Million for Charity.
What does running a category actually require of a newcomer?
- Pick the category by skill shape. Any% demands routing knowledge and execution of a handful of hard tricks; 100% demands stamina and consistency; glitchless demands clean fundamentals. The category shapes your practice, so choose by what you want to train.
- Read the rules page before running anything. Every board's rules document defines legal timing starts, banned glitches and verification requirements. Ten minutes there saves a rejected submission later.
- Expect the 80/20 of improvement. Most categories reward a small set of core techniques executed consistently; the visible top-end tricks are the last 20 percent of time saved and the first 200 percent of frustration.
Hardware note: emulator and original-console runs usually compete on separate boards or with platform tags, and input display plus unedited audio-video sync are standard verification requirements.
Timing conventions deserve one more sentence, because they trip up every newcomer. RTA (real-time attack) clocks the run as humans experience it, including load times on older hardware; loadless timing, common on PC platforms, subtracts loading screens with frame data; and in-game time follows the cartridge's own clock, which some games manipulate with save-and-quit tricks. Boards specify which convention is official, and a runner who trains on one convention and submits on another is the most common rejection in verification queues. Check the rules page for the timing start too — when the timer begins (input one, file select, cutscene end) is a community decision, and it is never obvious.
Which categories anchor the big events?
Marathon events like Games Done Quick curate for watchability, not raw difficulty: a 15-minute any% with two hard tricks usually beats a 4-hour 100% for the schedule. That curation feeds back into the community — categories that perform well on stream attract new runners, which is why any% boards stay the deepest talent pools. Speedrun Records events, European marathons and game-specific online summits cover the rest of the calendar, each with its own submission culture.
Verdict: start with your game's glitchless or easy any% category, read the rules document first, and treat the leaderboard as a ladder of routing rather than reflex alone. If you are choosing what to watch, any% shows a game's brokenness, 100% shows its design, and low% shows its community's obsession — three different sports wearing one cartridge. Should you run or should you watch? Run: every category board's mid-tier is friendlier than the record page makes it look.