Project Zomboid's Chinese-English review gap, re-measured

Steam’s appreviews endpoint for Project Zomboid (appid 108600), queried on 13 September 2026 with no language filter beyond Simplified Chinese or English, returns 24,988 positive and 4,046 negative Simplified Chinese reviews (86.1% positive, all-time) against 193,901 positive and 13,523 negative English reviews (93.5% positive, all-time) — a 7.4-point gap. That is the full population Steam itself scores as “Very Positive” in both languages, and it does not resemble the 44-point gap this article previously reported from a 100-review-per-language recent sample.

Why the number changed

The version of this article published on 2 September reported 84% English and 40% Simplified Chinese from the 100 most recent reviews in each language, then — in a 5 September correction — separated that score sample from a larger 800/200-review topic sample without being able to revalidate either one, because the underlying keyword rules and review-level snapshot were not recoverable. Neither problem exists with a query_summary pull: it is Valve’s own aggregate over every review Steam holds for that language, it needs no snapshot to reproduce, and re-running the same request returns the same numbers by construction. The 84%/40% figures were a real read of a 100-review window, but a 100-review window is small enough that one cluster of same-day negative reviews — a common effect right after a patch — can swing the score by dozens of points. The full-population number does not have that failure mode.

Both languages read “Very Positive” against Steam’s own scoring, all-time. The gap that survives — 7.4 points — is real and reproducible, but it is a fraction of what the earlier small sample implied.

What negative reviews are actually about right now

A query_summary tells you the split; it does not say what negative reviewers are complaining about. For that, the 200 most recent Simplified Chinese negative reviews (2026-08-20 to 2026-09-13) and the 199 most recent English negative reviews (2026-09-13 pull; the same request returned only 199, one short of 200) were pulled with review_type=negative&filter=recent and classified by whether the review text contains any of a small keyword set per topic — a mention-based count, not a mutually exclusive category, and two topics can both be true of one review.

The exact keyword lists used, matched as case-insensitive substrings against the full review text (a review counts toward a topic if any one keyword for that topic appears anywhere in it):

• Updates / version — Chinese: 更新, 版本, 补丁, 热修. English: update, patch, version, hotfix, b42, build 42
• Mods / Workshop — Chinese: 模组, 创意工坊, mod, MOD. English: mod, mods, workshop
• Multiplayer / servers — Chinese: 联机, 服务器, 多人, 组队. English: multiplayer, server, co-op, coop
• Performance — Chinese: 卡顿, 掉帧, 优化, 性能, 帧数, 闪退, 崩溃. English: fps, lag, crash, optimi, perform, stutter
• Translation — Chinese: 翻译, 汉化, 机翻, 语言. English: translat, localiz
• Price / refund — Chinese: 退款, 价格, 打折, 折扣. English: refund, price, expensive, overpriced

Substring matching means some of these overcount: the English “mod” keyword also matches “mode” or “modern,” and “perform” matches “performance” and “performing” alike. That inflates the raw counts above by an unmeasured amount rather than deflating them, so the true mention rate for each topic is probably lower than shown, not higher.

Mods/Workshop mentions run about 1.6x higher in the Chinese sample than the English one — a real, present gap, but smaller than the roughly 2x figure the earlier version of this article reported from an unrecoverable dataset. Translation shows the largest relative gap of any topic (9.5% against 0.5%), which the earlier version of this article also found but could not connect to the score gap; that remains true here.

What this sample does not show about the August API change

The 26 August hotfix that shipped 42.20.4 removed the loadstring and loadstream methods from the Lua modding API, breaking any mod that used them. The earlier per-language score comparison this article draws its own comparison against used a different sample entirely — the 100 most recent reviews of any type, not the negative-filtered pulls below — so the two articles are not measuring the same set of reviews. Searching both negative-review samples for the literal strings loadstring or loadstream returns zero matches in either language — not a low rate, zero. That does not mean the removal did not cause mod breakage reflected in these reviews; it means players describe broken mods without naming the removed method, in Chinese and English alike, so this sample cannot attribute any specific negative review to that one API change. The “mods/workshop” mention counts above are the ceiling on what this dataset can say about mod-related complaints; they are not evidence that the August change specifically drove the Chinese-English gap in that count.

Reading this before buying, or before quoting a score

• The all-time, all-language score is 93.8% positive (“Very Positive”) per Steam’s own totals; the 86.1%/93.5% split above is language-filtered, not the headline store score.
• A 100-review sample and a query_summary are answering different questions. The 100-review sample tells you about the last few reviews at whatever moment you pulled it; the query_summary tells you about every review Steam has logged for that filter. Neither is wrong — they measure different things, and a 100-review window can move by dozens of points between pulls.
• A mention count is not a cause. The topic table above says what negative reviewers write about, at roughly what rate, in two languages. It does not establish that any one patch produced any one gap.

Sources

• Steam appreviews endpoint, Simplified Chinese, query_summary — read 2026-09-13; 24,988 positive / 4,046 negative / 29,034 total, all-time
• Steam appreviews endpoint, English, query_summary — read 2026-09-13; 193,901 positive / 13,523 negative / 207,424 total, all-time
• Steam appreviews endpoint, all languages, query_summary — read 2026-09-13; 451,124 positive / 29,949 negative / 481,073 total, all-time
• Steam appreviews endpoint, Simplified Chinese, 200 most recent negative reviews — pulled 2026-09-13 in two 100-review pages by cursor; reviews dated 2026-08-20 to 2026-09-13
• Steam appreviews endpoint, English, 199 most recent negative reviews — pulled 2026-09-13 in two 100-review pages by cursor; reviews dated 2026-08-30 to 2026-09-13
• 42.20.4 STABLE & 42.19.2 UNSTABLE & 41.78.21 LEGACY Hotfixes Released — 2026-08-26, the loadstring/loadstream removal