The Ash Marshal is a genuinely fine encounter. Two phases, a moveset that rewards standing closer than feels safe, and a fourteen-frame minimum tell that sits right at the edge of our fairness threshold and holds because the animation is unambiguous. It is one of the six fights here we would recommend without qualification.
You meet it again ninety minutes later under a different name with forty per cent more health, and it is the same fight, and something in the game dies at that moment.
What counts as a repeat
We are careful about this, because variants can be legitimate — Ember Antler is built entirely from them. Our test is whether the encounter introduces at least one attack that requires a new response. If everything you learned in the first fight is sufficient for the second, it is a repeat.
By that test, nine of twenty-eight are repeats and seven of those nine differ only in health pool and particle colour. Two add a single attack, which is enough to keep them out of the category but not enough to make them interesting.
| Encounter | Original | Change | New responses |
|---|---|---|---|
| Cinder Marshal | Ash Marshal | +40% health, blue flame | None |
| Twin Wardens | Warden Of The Gate | Two at once | Positioning only |
| Hollow Knight-At-Arms | Knight-At-Arms | +55% health | None |
| Pale Houndmaster | Houndmaster | +1 attack | One |
| Sable Marshal | Ash Marshal | +70% health | None |
The six that are worth the price
The Ash Marshal on first meeting, the Gate Warden, the Drowned Choir, the Bell Sister, the Rider and the final encounter. All six have distinct movesets, minimum tells between fourteen and twenty-one frames, and arenas sized appropriately — none of the spatial failure that sinks Hangar Cluster.
Our attempt log shows one hundred and sixty-one of two hundred and eighty-nine attempts spent on these six. That is a reasonable game's worth of encounter design, and if it had shipped as a game with six bosses it would be sitting near nine.
Nobody has ever finished a game and wished it had padded its boss roster. The count on the box is for the people who have not played it yet.
What the repeats cost, measured
Our named-cause figure here is seventy-nine per cent, below the archive median of eighty-four. That is not because the tells are poor — the fights that are distinct read cleanly. It is because attention collapses during repeats. Loggers reported disengaging during the second and third Marshal fights and dying to attacks they had comfortably handled two hours earlier.
This is a real effect and worth naming: a repeated encounter does not merely fail to add anything, it actively degrades the player's engagement with the material around it.
| Measure | Sable Ring | Archive median |
|---|---|---|
| Shortest tell | 14 frames | 18 frames |
| Attempts to first clean phase | 12 | 14 |
| Camera or geometry deaths | 14% | 9% |
| Named-cause deaths | 79% | 84% |
| Reused encounters | 9 of 28 | 3 of 14 |
The twin fight deserves separate blame
Two Gate Wardens at once, in the arena built for one. Neither has been redesigned to account for the other, so their attacks overlap in ways nobody choreographed, and the camera cannot hold both. It produced twenty-two of our attempts and eleven of our camera deaths, and it is the single worst encounter in the game.
The short version
- Twenty-eight advertised bosses, nineteen distinct movesets.
- Seven repeats differ only in health pool and particle colour.
- Six genuinely excellent encounters, one hundred and sixty-one of our attempts.
- Named-cause deaths fall to seventy-nine per cent, driven by disengagement.
- A doubled encounter in a single-boss arena is the worst fight here.
Verdict
Seven point four. Six fights this desk would defend in any company, buried in a roster padded to reach a number on a box. Play it for the Marshal, the Bell Sister and the Rider, and treat every recoloured repeat as the interval it is.