The Needle'sVerdict

So the ordering below is closer to a measurement than an opinion. Read it by the bars rather than the ranks: the bar shows how the whole vote split, and that shape carries more than any average can.

That level of agreement means the survey was enormously oversampled for the question it asked. Working backwards: how many respondents would you actually have needed?

The honest version of the ranking is therefore slightly shorter than 46. Testing each neighbouring pair on just the players who fought both, 15 pairs fail to separate. Collapsing those leaves 31 genuinely distinguishable positions, with twelve tied blocks:

A five-point scale can only measure what fits inside it, and seven bosses broke out of the top. First Sinner was scored 5 out of 5 by 87.4% of everyone who fought it. Skarrsinger Karmelita by 84.6%, Lost Lace by 76.2%, Cogwork Dancers by 71.1%. Counting everyone who said 4 or 5, First Sinner reaches 96.9% — essentially the whole playerbase.

This is a ceiling effect, and it changes how the rest of the table should be read. The gap between First Sinner (4.82) and Widow (4.53) looks trivial next to the gap between Widow and Bell Beast (3.19). It almost certainly isn't. Once four-fifths of your sample has chosen the maximum, the instrument has stopped measuring — the real distance between Silksong's hardest fights is compressed into three tenths of a point because there was nowhere else for those players to put their answer.

The same wall exists at the bottom, less dramatically: 85.0% found Plasmified Zango easy, and 77.9% said the same of Palestag.

Re-scoring every rating relative to that respondent's own average — which strips the harsh-grader effect out entirely — changes the rank order of exactly zero bosses. But agreement is not uniform, and where it breaks down is the interesting part.

That flatness is not what "medium difficulty" looks like. A genuinely medium boss looks like Bell Beast, where 54.1% of players chose exactly 3 and the divided score is the lowest in the dataset outside the extremes. Groal isn't medium. Groal is two different fights depending on who you ask — and it turns out you can find the two groups.

Ignore how hard each player found the game overall, and group people purely by the shape of their answers — which bosses they found relatively harder than the rest. Restricting this to the 9,902 respondents who fought at least 44 of the 46 bosses, so that missing answers can't drive the result, the playerbase falls into two clean groups of 5,978 and 3,924.

The two camps disagree about almost nothing. Across 43 of the 46 bosses their scores sit within about half a point. Then there are three fights.

So the single largest genuine disagreement in Silksong's playerbase is about the two Savage Beastfly encounters and Groal the Great — three fights that roughly 60% of players found trivial and 40% found middling. For Savage Beastfly I the two camps average 1.70 and 2.98. For the Far Fields rematch, 1.59 and 2.92.

So what do these three fights have in common? Cross-referencing a published boss guide gives an answer that fits almost too neatly: each one has a documented way to trivialise it. Groal can be fought from under the lower-right platform, where none of his attacks reach. Savage Beastfly can be lined up so that its own vertical charge kills the allies it summons. Both are described in strategy guides as cheese.

The survey asked nothing about tools, routes or tactics, so this remains an inference from an external source rather than a measured result. But it is a testable one: a follow-up poll that asked how people beat Groal would settle it in one question.

Split respondents by how much of the game they got through — 45 or more bosses fought versus 40 or fewer — and compare their scores on the 32 bosses that almost nobody skipped, so both groups are rating the same fights.

The pattern is that the gap opens exactly where players' circumstances diverge. Where the game funnels everyone through the same door at the same power level, the scores converge. Where it doesn't — the optional, out-of-the-way, late fights — they split. Groal again leads, at half a point.

This is an interpretation of the pattern rather than something the survey measured; it did not ask about routes, tools or build. An alternative reading is simply that thorough players remember more of the struggle, having spent longer in each fight.

That inverts the usual expectation for a Metroidvania, where optional content is where the real challenge hides. In Silksong the hardest fights are mostly on the critical path, and the optional roster is where the walk-throughs live: Plasmified Zango, Palestag, Broodmother, both Savage Beastflies, Voltvyrm — all optional, all rated below 2.2.

So the shape of Silksong's difficulty is: a demanding critical path, a long tail of easy optional content, and a small number of hidden fights that are harder than anything the game requires. And this audience found nearly all of them.

Skip rates confirm it from the other direction. Fourteen bosses were skipped by more than 5% of respondents, and they are overwhelmingly the easy optional ones — skip rate correlates negatively with difficulty (−0.25). Silksong gates its hidden content behind exploration, not skill.

Read the other way, this is a striking measure of the audience: nearly a third of respondents had completed Silksong at least once and then started again on permadeath, far enough in to reach a secret boss in Moss Grotto. Its score of 2.28 comes from the most hardened subgroup in the entire sample, which is exactly why it should not be compared directly with the rest of the table.

A useful test of whether a survey measured anything real: does it recover structure nobody told it about? Strip out each player's personal harshness, then correlate every pair of bosses. The strongest relationships in the entire 46×46 matrix are, without exception, the same boss fought twice.

Below the repeat encounters, a second cluster appears with no shared name at all: First Sinner and Skarrsinger Karmelita (0.43), Lost Lace and Karmelita (0.38), First Sinner and Widow (0.38). That is the endgame gauntlet showing up as a statistical object — the fights that filter the same players out.

Then there is the response curve. 78.3% of all 16,361 responses arrived in the first three days, 7,625 on 10 December alone. This is one viral spike, which means the sample is a snapshot of one community at one moment rather than a rolling cross-section of players.

One more wrinkle sits in the timestamps. The 1,324 people who answered more than a week after launch did not score things the way the viral-day crowd did — they pushed the hard bosses higher and the easy bosses lower, spreading their answers wider rather than compressing them. The relationship between a boss's difficulty and how much late respondents shifted it is +0.44. Late arrivals polarise. Whether that is a different kind of player or a different kind of memory, the survey cannot say.

Silksong does not ramp. It oscillates, and violently: Widow at position 10 scores 4.53, then Moss Mother II at 11 scores 2.39 and Savage Beastfly II at 12 scores 2.06. First Sinner, the hardest fight in the game, sits at position 26 of 48 — the middle. Broodmother, the fourth easiest, comes immediately after it.

What does change across the game is the spread. Act 1 covers 2.48 points from its easiest boss to its hardest; Act 2 covers 2.92; Act 3 covers 3.13, holding both Skarrsinger Karmelita at 4.76 and Plasmified Zango at 1.63. Average difficulty drifts only slightly upward by act — 3.14, 3.32, 3.34 — while the share of maximum scores climbs from 39.3% to 46.9% to 47.2%.

Silksong doesn't get harder so much as it gets wider. The late game raises its ceiling without raising its floor, and the result is not mounting pressure but sharper and sharper contrast between the walls and the walk-throughs. A player's memory of "the difficulty spike" is really a memory of one of these teeth.

Hard % and Easy % are the shares scoring 4–5 and 1–2. Divided is the 0–100 split score. Rank range is the 95% confidence range on the position from 1,000 resamples. Vet gap is the completionist difference, blank where the boss was skipped too often to compare fairly.