The first hour of Canyon Shell is one of the best things in this issue. A creature the size of a hillside, plated like a woodlouse, and the entire fight is a problem of access: which plate can you reach, what does removing it expose, and what does the creature start doing once it knows you have found a way in.
The answer to the last question is that it stops telegraphing.
The mechanic that undoes itself
As plating is removed the creature becomes faster. That is a reasonable design instinct — it is wounded, it is desperate, the fight should escalate. The implementation is that the same attack animations play at a higher rate, which means the wind-ups compress along with everything else.
The slam is twenty-six frames when the creature is intact. Fully exposed it is eight. Our archive shows that below nine frames, players stop describing a fight as hard and start describing it as broken, and they do so at almost exactly the same attempt count regardless of damage output. We reached that point at attempt nineteen.
| Plates removed | Animation rate | Slam wind-up | Our named-cause rate |
|---|---|---|---|
| None | 1.00× | 26 frames | 91% |
| Two | 1.35× | 19 frames | 77% |
| Four | 2.00× | 13 frames | 54% |
| All six | 3.25× | 8 frames | 18% |
That final row is the review. In the state the fight spends its last third in, we could explain fewer than one death in five.
Rate-scaling is not difficulty design
Speeding up an animation is the cheapest possible way to make a fight harder, and it is the one that most reliably converts an encounter into a memorisation exercise. Nothing new is being asked of the player: the same four attacks arrive, in the same order, from the same silhouettes, only now the silhouettes are unreadable.
Compare Spire Of Glass, whose shortest tell is thirteen frames and which never once felt unfair. Thirteen frames of a whole visible human body carries information. Eight frames of a limb on a creature the size of a hill carries none, and the difference is not the number.
Escalation should add something. Multiplying the animation rate subtracts the tell and calls the result a challenge.
The other four bosses
Ordinary, and mercifully free of rate scaling. All four sit between eighteen and twenty-four frames at their shortest and produce named-cause figures above eighty-five per cent. They are pleasant, forgettable hunts, and they demonstrate that the studio knows how to build a legible fight when it is not trying to escalate one.
The camera is also a problem, though a smaller one: nineteen per cent of our deaths came from geometry, largely because the creature's exposed underside is where you want to be and the canyon walls are where the camera goes instead.
| Measure | Canyon Shell | Archive median |
|---|---|---|
| Shortest tell | 8 frames | 18 frames |
| Attempts to first clean phase | 21 | 14 |
| Camera or geometry deaths | 19% | 9% |
| Named-cause deaths | 52% | 84% |
| Rate-scaled encounters | 1 of 5 | 2 of 14 |
What we would change
Cap the animation rate at roughly one point six times, which holds the slam at around sixteen frames, and add a distinct audio cue to the exposed-state slam. That is all. The idea underneath — that a boss is a structure you take apart — is strong enough to carry a whole game, and it does carry the first hour of this one.
The short version
- Excellent premise: a creature dismantled plate by plate.
- Animation rate scales to three and a quarter times as plates are removed.
- Slam wind-up compresses from twenty-six frames to eight.
- In the final state we could explain fewer than one death in five.
- The other four bosses are ordinary and entirely legible.
Verdict
Six point eight. One of the best encounter ideas of the year, ruined in its final third by the laziest available difficulty lever. We would return to this in a heartbeat if the rate scaling were capped, and we suspect the fight underneath is worth eight and a half.