Findings that hold across the archive
The Eight-Frame Cliff
Across four hundred and eleven encounters, complaints of unfairness do not rise smoothly as wind-ups shorten. They jump, sharply, somewhere between nine and eleven frames — and the jump happens at the same attempt count regardless of how much damage the attack does.
What Second Phases Do With Your Memory
Just over half the archive discards the first phase's moveset when the second begins. We compared retention against named-cause death rates and found the relationship is not subtle: fights that keep their vocabulary are explained by their players, and fights that replace it are not.
Measurements we are still collecting
Arena width against widest reach
Our working threshold is seventy per cent: where a boss's widest attack covers more than that proportion of the arena, positioning stops mattering and the fight becomes a single dodge timing. Hangar Cluster put five of six encounters above it. We need roughly a hundred more measurements before publishing.
Run-back distance and abandonment
How far a checkpoint sits from a boss door, and at what distance players stop retrying. Early figures suggest abandonment climbs sharply past forty seconds, but the sample is contaminated by fights people were abandoning for other reasons.
Rate scaling versus added attacks
Whether escalation by animation speed produces different player descriptions than escalation by new moves. Canyon Shell is the clearest case we have logged, at three and a quarter times, but clean comparisons are rare.
Why we publish the notes separately from the reviews
A review has to answer whether one game is worth your hours. A note has to answer whether a pattern is real, which requires a sample large enough that no single game can move it. Mixing the two produces reviews that argue with data instead of describing a fight, and notes that generalise from whatever came out last month.
Every figure in these notes is measured from our own capture footage, on retail builds, using the procedure in the method. Where a sample is too small to support a claim, we say so and leave the note open rather than publishing a number we would not defend.
- Published Sample above two hundred, effect survives removing any one game
- Provisional Effect visible, sample below two hundred
- Open Collecting, no claim made
- Withdrawn Effect vanished as the sample grew
- Revised Threshold moved after new measurements