Speedrunning on cloud — the rules conversation we should be having

Speedrun.com leaderboards distinguish runs by platform, version, category, and various game-specific rules. Some games have separate leaderboards for emulator runs vs original hardware. Some require specific firmware versions. Some disallow turbo controllers.

The community has spent years codifying what makes a run legitimate. The codification is detailed and serious — top runners' world record runs are reviewed frame-by-frame for evidence of tool-assisted manipulation.

Cloud gaming sessions add 30-60 ms of input latency on top of the game's local input latency. For most speedruns this is a disadvantage (the runner has slightly less precise input) but in some specific cases it's neutral or beneficial.

Cases where cloud might help: games where the optimal route involves frame-perfect inputs that benefit from a consistent latency floor. Cloud's latency variance is sometimes lower than a local rig's because the cloud GPU's frame timing is more consistent than a thermally-throttled local GPU.

Cases where cloud likely hurts: most speedruns. The input lag is a net negative for the same reason it's a net negative in competitive shooters.

Should cloud-played runs be allowed on the main leaderboard for a game? The community hasn't decided. Some games have explicitly disallowed cloud runs because the latency variance makes verification harder. Others allow them and treat them as equivalent to local runs.

The verification angle is the trickier piece. A speedrun video showing the game running in a cloud client is harder to authenticate than a video showing a locally-running game. The cloud client could in principle be running modified game logic upstream of the video stream, and the runner can't prove it isn't.

Some leaderboard moderators have moved cloud-played runs to a separate category. This is the cleanest answer but creates a two-tier system that hasn't been universally adopted.

GeForce Now-played runs are increasingly allowed on mainstream leaderboards for non-marquee categories. The cloud-vs-local performance difference is small enough on most non-frame-perfect categories that the community has accepted them.

Game Pass Cloud runs are rarer because the title catalogue doesn't include most popular speedrun games. Where they exist (Forza, Halo), the community has been cautiously inclusive.

Console cloud streaming (PS Plus Premium cloud sessions) is essentially absent from speedrun leaderboards because the verification overhead doesn't seem worth the small audience.

An explicit 'cloud streaming' category on major leaderboards. Treats cloud-played runs as a distinct category, mirroring the way emulator runs are separated from original hardware. The cloud category is then internally consistent (cloud vs cloud) and the verification overhead is manageable.

Better video-evidence standards for cloud-played runs. The cloud client overlay should be visible in the run video. Cloud session metadata (timestamp, region) should be captured. The verification can be tighter if the standards exist.

A clearer position on cloud-played 'any%' categories where input latency matters less. These could plausibly join the main leaderboard with proper categorisation.

Cloud gaming is going to expand the speedrunning audience by making more games accessible to more people. The audience growth is good for the speedrun community generally.

But the leaderboard infrastructure has to keep up. Without explicit cloud categorisation, the community ends up in awkward 'are cloud runs legitimate' disputes that distract from the actual running. The categorisation is a small policy decision with a larger community-health benefit.

I'd hope to see Speedrun.com adopt a standardised cloud-category convention in 2026. The community is large enough now that the policy decision should be made centrally rather than re-litigated per-game.