Why publish a methodology
Our guides quote numbers such as average scores, hits per run and how much dashing helps. Those numbers come from running the real game engine, not from estimates, and this page explains how so you can judge them yourself. We update both the numbers and this page whenever the game rules change.
The setup
- Engine: the same deterministic engine the game and server use (engine version 3, measured on 2026-10-22).
- Arenas: 200 fixed seeds for each of the seven twists, identical across every configuration so comparisons are like-for-like.
- Test player: a deliberately simple script that steers toward the nearest shard (or a nearby power-up) with no route planning. One version never dashes; the other dashes whenever a hazard comes within 45 units.
- Measured: score, median and top-10% score, hits, run length, shards, power-ups, near misses, and the share of shard points earned at combo ×5 or higher.
Limits
The test player is not a human. It reacts instantly but never plans, so it is better than a beginner and worse than a skilled player who routes through shard clusters. Treat its numbers as a consistent benchmark, not as a prediction of your score. As one reference point, it averages 8,025 on the classic rules when dashing.
Leaderboard scores from real players are not used in these figures. If we publish analysis of real play in future, it will be aggregated and labelled as such.
Reproducing the numbers
The measurement script lives alongside the game's source and regenerates every figure in one command, so the guides can never silently drift away from the rules. Questions or corrections are welcome via the contact page; see also our editorial policy.