Mortal Shell II — The Simplified Chinese Steam Rating Gap

Analysis date: 21 August 2026 Basis: all 5,951 Steam reviews, classified and counted Question: English reviewers rate Mortal Shell II 86% positive. Simplified Chinese reviewers rate it 57%. Why?

Answer in brief

Both audiences report largely the same faults. Six of the eleven complaint themes appear at near-identical rates in Chinese and English reviews — bugs, boss fights, checkpoints, camera, enemy variety, and the sense that the demo misrepresented the release.

The gap comes from a specific cluster of four themes where Chinese incidence runs two to four times higher, and those four are not a random assortment. They are all about the same thing: the rate at which the game wears you down.

Enemy density, map structure, upgrade scarcity, damage scaling. Take them together and they describe one experience: dying often, to crowds, far from a checkpoint, in a place you cannot navigate, without the resources to get stronger. Chinese reviewers name that combination far more than English ones — and they have a word for it.

The second finding narrows the cause. Traditional Chinese reviewers sit at 78% — twenty-two points above Simplified, nearer English than Chinese. Same characters, same build. Whatever is happening is specific to the mainland audience, not to reading Chinese.

The third finding locates it in time. The gap ran 22 points at launch and widened to 33 on 19 August. The reviewers who moved it had a median of 12 hours played, up from 7. This is a problem in the back half of the game, not a first impression.

The numbers

Steam’s own language-filtered figures, 21 August:

Positive rate by language across the full corpus, worst to best:

Simplified Chinese is the floor by a wide margin — twelve points below the next language. Russian at 86% matches English, which rules out any “West versus the rest” reading.

What players actually say

Four terms carry most of the Chinese criticism. They are worth knowing because they are more precise than their English equivalents.

堆怪 — “mob stacking.” Not “there are many enemies.” It accuses the game of using enemy quantity as a substitute for encounter design. It appears in 26% of Simplified Chinese negative reviews.

篝火 — “bonfire,” the checkpoint. Appears in 20%. Almost always in the form of there being too few, too far apart.

数值 — “the numbers,” meaning stat tuning. Appears in 21%, usually asserting that enemy damage and health were scaled up rather than encounters being redesigned.

恶意 — “malice” or “hostile design.” The strongest term: the accusation that the game is deliberately wasting your time. Peaked at 11% of Chinese negatives and has since fallen to 4%.

The most-upvoted Chinese negative review — 84 players marked it helpful, 20 hours played — contains four of the eleven themes at once:

通关后依旧是垃圾游戏，只能感受到满满的制作组的恶意，全图没几个篝火，严重的堆怪，数值及其崩坏 […] 只能说被demo骗完了

“Even after finishing it, it’s still a garbage game. All I can feel is the developers’ hostility — barely any beacons on the whole map, severe mob stacking, completely broken stat scaling. […] All I can say is the demo fooled me completely.”

Another, with 41 upvotes, sets out the chain in order:

堆怪严重小怪数值又高，篝火点也很少 […] 倒地没有无敌一套连招躺在地上被连死，重新跑图就是三百六十五里路

“Severe mob stacking and trash enemies with inflated stats, and very few beacons. […] Once you are knocked down the game gives you no protection while you get up, so they simply finish you off on the floor — and then the walk back is a three-hundred-mile trek.”

On map design, the recurring charge is illegibility rather than size:

地图一团乱麻毫无辨识度，可以说是迷宫跑图模拟器

“The map is a tangled mess with no landmarks — you could call it a maze-jogging simulator.”

English reviewers reach the same conclusions in a different register — more often separating the game from its problems:

“There’s a really good game here buried under poor enemy design and balance.” — 22 upvotes, 8 hours

“A good game dragged down by mind boggling design decisions.” — 26 hours

“I have as many crashes as I do hours in this game.” — 16 hours

Finding 1 — the shared complaints, and the four that are not

All eleven themes, share of each language’s negative reviews:

Base: 579 Chinese and 375 English negative reviews. A review arguing two things is counted under both, so columns sum past 100%.

The shared half of this table is the important context. Bugs, bosses, checkpoints, the camera, enemy variety, the demo grievance — all raised at comparable rates by both audiences. This is not two populations playing different games or holding incompatible standards.

The divergence is concentrated and coherent. The four Chinese-skewed themes are the four that describe attrition: how many enemies at once, how far back the checkpoint is, how hard they hit, and whether you can get stronger. English reviewers hit those too — 8–10% each — but for Chinese reviewers they are the dominant register.

The one English-leaning theme, parry and hit detection at 1.6×, points the same way from the other side: English reviewers are more likely to frame the difficulty as a mechanical reliability problem, Chinese reviewers as a design intent problem.

Finding 2 — splitting Chinese in two narrows the search

Traditional Chinese works as a control group: it shares the written language with Simplified but not the audience. Two groups reading the same characters, playing the same build, land twenty-two points apart — with Traditional closer to English.

That eliminates the explanations operating at the level of language: comprehension, translation quality, localisation. It does not eliminate culture, because Simplified and Traditional readers are also different markets — mainland China versus Taiwan, Hong Kong and Macau — with different pricing, genre habits and platform norms. What the split does is narrow the question from “why do Chinese players rate this lower” to “why the mainland audience specifically.”

Where the two part company:

Mob stacking is the one complaint both raise identically. Everything else in the attrition cluster is close to absent from Traditional Chinese reviews.

Treat this table as indicative: it rests on roughly 22 Traditional Chinese negative reviews. The 22-point headline gap is solid on 101 reviews; the term-level split is not.

Finding 3 — the gap widened, and the second wave had played longer

English drifted down five points across the launch window. Simplified Chinese fell seventeen, nearly all of it on one day.

The reviews from the 19th onward differ from those before:

The later wave had played nearly twice as long, wrote more briefly, and reached less for moral framing. Note what that means: the vocabulary of hostility was a feature of the early response and has more than halved, while the score kept falling. The people who moved the number most were not the angriest — they were the deepest in.

Chinese negative reviewers play longer than English ones before writing: median 10 hours against 7, with 24% past 20 hours against 18%. They are not bouncing off the opening.

The mechanism

Putting the cluster and the timing together gives a single sequence, which the reviews describe directly:

a space you cannot read → a crowd you cannot fight one at a time → a hit that takes most of your health → knocked down and finished before you can stand up → a long walk back to where you died → too few resources to come back stronger → repeat

Each link is one of the four skewed themes. No single link is unique to Chinese reviews; English reviewers report every one of them. What differs is how often the whole chain gets named as the reason for a thumbs-down, and how much further into the game it starts to dominate.

Steam then compresses this into one bit. A player who finds the game good but exhausting and a player who finds it excellent both leave the same mark, and so do a player who is mildly irritated and one who wants a refund. English reviews are full of the “good game, bad decisions” construction that still resolves to Recommended. That framing appears far less in Chinese reviews, and the binary vote makes the difference in resolution fully visible in the headline percentage.

Method

Themes are assigned by multilingual keyword sets built from the measured vocabulary of the negative reviews — the most frequent n-grams per language — rather than from guessed terms. A review counts as substantive if it is a thumbs-down, long enough to contain an argument, from someone with at least an hour played, and points at something specific rather than only venting.

All theme incidence is normalised within language: the share of that language’s negative reviews raising the theme. Without this, Chinese would appear to dominate everything simply because it has more negative reviews.

Every figure is the length of a list of review IDs and can be recounted from the cache. The full classified set, with each review’s text, playtime and link, is on the Player-reported issues page.

Limitations

Scope is Steam only. No Reddit, no social, no press. The cross-source counts elsewhere on the site are wider and not comparable with these.

Keyword assignment under-counts. A complaint phrased without the expected vocabulary is missed; 48 substantive negative reviews matched no theme at all. An earlier version of this analysis was materially wrong for exactly this reason: the matcher tested Chinese reviews only against Chinese terms, and did so case-sensitively. Chinese reviewers use Latin-script loanwords heavily — “bug” appears in 144 of 579 Chinese negatives, “boss” in 179, “demo” in 90 — so several themes were undercounted for Chinese. Before the fix, the demo complaint looked five times more English than Chinese; it is actually near-identical. Any keyword method carries more of this kind of error than it can see.

Traditional Chinese is a small sample. About 22 negative reviews behind the term-level table.

Steam’s percentages differ from ours by a point, because Steam’s base includes reviews the public API does not return.

Attribution, not causation. Counting what people say identifies where attention concentrates. It cannot establish that the attrition cluster caused the votes, only that it dominates the stated reasons.

Relationship to the earlier report

An initial analysis on 19 August reached a similar conclusion from a reading of representative reviews rather than a count: that both audiences report the same problems, and that Chinese reviewers judge their combination more harshly under the label 恶意. The measurement largely supports the first half — six of eleven themes are shared — and adds the part that reading could not: that the divergence is concentrated in four attrition themes at 2–4× incidence, that Simplified and Traditional Chinese split by 22 points, and that 恶意 was fading even as the score fell. That report is superseded by this one; nothing in it needs to be read first.

Conclusion

Chinese and English reviewers agree about most of what is wrong with Mortal Shell II. They diverge sharply on one axis: how much of the criticism is about being ground down.

Enemy density, illegible maps, scarce upgrades and late-game damage scaling account for the gap, at two to four times the English rate, and they compound into a single loop that the reviews describe step by step. The problem deepens as players get further in — the reviewers who pushed the score down hardest had twelve hours in, not two.

The most useful evidence is internal to Chinese. Traditional Chinese players, reading the same characters, sit twenty-two points higher and barely mention the maps, the checkpoints or the numbers. The question is not why Chinese players are harsher. It is what the mainland audience is running into that others are not.