Steam reviews: issues & requests

FindingThe complaint is still not that the bosses are too hard — it is that most of them do not register as fights at all, and the ones that do are decided by a single attack rather than by play. What the new evidence changes is where the weight sits: the endgame, and specifically the final boss and Monolith, is now the centre of the theme rather than its last chapter.

• 34% of the reviews raising this still recommend the game
• Traditional Chinese reviewers raise it 1.6× less often (33%)

• Correction to the previous Finding: the design credit no longer holds. The highest-upvote new review says the bosses “all kinda suck”, the warmest new voice manages only “bosses aren't ideal, but they aren't awful either”, and the praise this theme rested on (“The bosses are generally well designed”, “I especially enjoy the boss fights”) now belongs entirely to reviews from the previous pass.
• That most bosses do not land recurs unchanged, and the new Russian review supplies the mechanism the English reviews only described: “former mini-bosses became full-fledged basic mobs”, packed two and three to a room, while story bosses get “reused in the open world”. The English line still states the shared point most precisely — “The trash mobs are the bosses in this game”.
• The fights that do land are still decided by one attack, but the new evidence sharpens what that means — not short fights, long ones with a fatal asymmetry: “damage from a poke on a boss 1/200 of the hp bar, damage from a perfect parry 2/200 … a minimal mistake will most likely send you to respawn”. The specific mechanisms recur as before: the first boss's 180-degree tracking command grab that “100 to 0's your hp”, grabs where “you almost always have less than a second to react”, and the Chinese parry-recovery lock that makes the second hit of a string neither parryable nor dodgeable.
• The final area and final boss now account for most of the new material and recur across languages: Monolith is named by English (“Late game boss sponges … looking at you Monolith and Zyem”) and Russian reviewers alike, a Chinese reviewer calls the final boss a sponge that force-feeds damage, a Japanese one quit there, and a Russian calls it “the most pathetic and shitty final boss of soulslikes I have ever met”.
• Boss-fight bugs have moved from a technical footnote to part of the complaint itself: a grab with a broken hitbox, a slam that “literally teleports you a kilometre from the boss” on both dodge and hit, a status effect the player cannot distinguish from “a critical bug that has been live for a week”, alongside the older bosses “yeeting me into the sun” and fights that can be skipped outright. The charge in these sentences is not difficulty but absence of testing — “DID YOU EVEN TEST THE LAST BOSS?”
• The missing-checkpoint complaint holds exactly as previously written and nothing new renews it: every Stake-of-Marika comparison is still English, and the 29-helpful Russian recommender still reports a bonfire next to the main boss. It remains attached to specific long dungeons and their runbacks — “add a one time heal restore and checkpoint in some of the longer dungeons” — not to bosses generally.
• The mini-boss-demotion and one-shot-economy framing is reported at a materially higher rate in Russian than in English or Chinese, where the same fights are described as forgettable or as runback problems instead.

The +315 mentions arrived at an essentially unchanged recommend share (33.3% → 33.8%) and the new sentences come from only three reviewers — two Russian, one English — so most of the growth is the existing complaint arriving late.

FindingThe complaint is not frame rate or polish but faults that land on the recovery loop — the crash or hang at a load transition, the elevator that will not come back up, the gloom that drops inside geometry — because each one makes you redo work that a death alone would not have cost. What the new arrivals add is a second exit-less state: not stuck in the map, but stuck in your own character, with quit-and-reload the only way out.

• 41% of the reviews raising this still recommend the game — a majority-adjacent share, which makes it a tuning complaint rather than a dealbreaker
• French reviewers raise it 1.4× more often than average (60% of their complaints)

• The highest-recurrence fault remains the wedge with no exit but the title screen, and the new arrivals widen it from geometry to input state: “the game prevents you from interacting with anything so you need to die or quit the game and reload it back up”, a bug where the player “couldn't attack until I restarted”, and a lock-on you “cant get out” of. Chinese files the same shape as 卡死 with 小退 — quit to menu — named as the fix, so quitting is the normal verb in both languages because the game offers no alternative.
• The crash still sits on the load transition and the new reviews locate it more precisely at beacon travel, where the game “может зависнуть на несколько минут и вылететь” — a multi-minute hang first, then the crash. Death respawn and continue-from-main-menu are still named alongside it, and nothing in the new arrivals reads as fixed.
• The gloom is destroyed by geometry as often as by a crash, and Russian arrivals repeat it directly — souls suffering “провала … на полу под текстуры”. This matters because it takes the loss out of the crash's hands; the run is wasted even when the game never quits, and Chinese reports the same loss at roughly half the English rate, so the difference is emphasis, not presence.
• Elevators are still the single most-named object across English, Chinese and Russian: you fall under them, the lever gives no signal whether it is working, and one player was “killed off a lift, that glitched into existence in front of me”. The object recurs because it sits on the runback, so a broken one blocks recovery rather than merely annoying.
• Correction to the last pass: the “worse since the hotfix” reading does not survive the new arrivals, none of which repeat it, while two recommenders report the opposite — “performance stabilizes after that, saves remain intact, and I experienced no crashes during combat” and “Performance was solid for me”. What recurs instead is that nothing got fixed, most pointedly from a 38-hour Russian unsure whether a last-boss effect is intended or a critical bug that “уже неделю висит”.
• The back-half concentration holds and is the clearest pattern in the new text: “The early to mid game is really fun … the late game areas are just miserable”, alongside “Feels like they rushed the entire last area” and enemy moves that “get more buggy the further you get”. The clone and summon-targeting cluster is still present but gained nothing new this window, so the lockouts are the only fault shape genuinely accelerating.

The 1355 is inflated — literal insects (“a bug with a human face”), the Chinese 无敌 i-frame balance talk, and balance gripes filed next to genuine defects all land in the count, so read the +218 as wider coverage rather than escalation.

FindingThe complaint is still not the number of enemies but that ordinary enemies were given boss-grade tools — one-shot damage, infinite poise, perfect tracking, grabs — while the player stays fragile and animation-locked. The new reviews sharpen the target twice: the lever players object to is the tuning numbers and the placement decision, and the cost is measured in the distance back from the last checkpoint.

• 22% of the reviews raising this still recommend the game — a low share, so this one tends to come with a thumbs-down
• Simplified Chinese reviewers raise it 2.4× more often than average (64% of their complaints)
• Korean reviewers raise it 2.3× more often than average (63% of their complaints)
• Traditional Chinese reviewers raise it 2.1× more often than average (58% of their complaints)
• English reviewers raise it 3.0× less often (9%)
• Russian reviewers raise it 3.7× less often (7%)

• The asymmetry, not the crowd, remains what players describe: most mobs have “a attack that has a VERY quick startup” and one flinch “makes it very easy for even the lowest of trash mobs to stun lock you to death”, while “my enormous armored murder-machine reacts to every enemy poke like he's been hit by a freight train, while most of the enemies tank your swings like nothing.” New reviews restate the same thing as poise — “most basic enemies have infinite poise during their attacks to the point they can never be interrupted.”
• Correction to the previous Finding: grabs are no longer the single most recurring boss-grade tool. They still recur and remain the sharpest case — one grab empties the bar, two or three enemies throw them at once — but the new material names the wider class more often, “almost every enemy have a one shot move and perfect tracking”, and that package on ordinary enemies is the narrowest fixable item.
• Checkpoint distance has grown from a consequence into the co-equal half of the complaint, and it now recurs in every language in this pass: German pairs the two directly — “Viele Gegner können dich innerhalb weniger Treffer töten, gleichzeitig sind die Abstände zwischen den Checkpoints teilweise unangenehm groß” — while tchinese objects to too few beacons and English reports runbacks of “multiple minutes”. One reviewer states the trade most precisely: “If you want to make your game feel like God of War, where every enemy is bulky and poses a challenge, you can't have your last checkpoint be miles away.”
• Placement is now named as the design lever itself rather than a side effect — “mehrere Elitegegner gleichzeitig auf den Spieler loszulassen, empfinde ich einfach als keine gute Lösung, um den Schwierigkeitsgrad zu erhöhen” — and the same reviewer sees difficulty raised through “Gegnermengen, hohen Schaden und teilweise fragwürdiges Gegner-Placement”. The older placement complaints hold unchanged: ambushes respawned into cleared rooms, blind-angle ganks, far-reaching aggro, and rooms too small to separate a group.
• The thin roster behind the multiplier holds: enemies met in the first two hours are still there in the final area with tripled health and damage, and a new German reviewer confirms variety falls off from mid-game onward. Scaling is erratic rather than rising — one enemy “deals 1 damage” while the next two-shots — and clearing a hard group pays “幾百魂完全無意義”, so players marathon past instead of fighting.
• Chinese reviews are 613 of the 865 (plus 28 tchinese) and still treat 堆怪 and 数值 as named anti-patterns rather than described experiences, where English describes mechanics; the new Chinese material adds the consequence that the tuning drives players off the game's own parry-and-execute system — “它的数值就在明确的逼迫你放弃交互打法” — into mutual one-shotting or simply running past.

The 865 counts sentences inside longer late-game complaints rather than reviews about stacking alone, and 22% still recommend — including a 52-hour platinum reviewer who makes exactly this complaint.

FindingThe parry itself is still liked; what players attack is that it has become the only answer to attacks that track them, while being the option that locks them in place, reads unreliably, and pays the least. The window is not really the complaint — the animation lockout after a successful parry and the gap between what an attack looks like and where it hits are.

• 33% of the reviews raising this still recommend the game
• German reviewers raise it 2.3× less often (10%)

• The parry is still what people defend, including reviewers who recommend the game — a Russian one calls the parries pleasant and the dropped stamina an interesting choice, and a critic states the split precisely: “The base parry feels good, but the perfect dodge window is very tight and inconsistent, even with enemies that I know the attack pattern of.”
• The top account in the theme (443 helpful, Chinese) remains the core mechanical complaint and is unchanged: parry the first hit of a dense combo and the recovery locks you out of both a second parry and a dodge, so the follow-up is guaranteed. New Russian and English reviews report it independently — “Ты никуда не денешься и не выйдешь из ситуации уклоном, потому что персонаж просто застревает” (you can't get out of the situation with a dodge, the character simply gets stuck).
• Sharpened this pass: dodging is not the escape hatch the previous Finding implied, because attacks reorient onto you regardless of position, so parrying becomes compulsory rather than a risk players choose. Three new English reviews land on the same clause — enemies “hard pivot 180 degrees mid swing”, lock on “like a heat seeking missile”, and both add unless you parry.
• Also new, and the clearest reason the reads fail: the visual and the hitbox are desynced, not merely mistimed. One reviewer recorded and reviewed his own footage to find “so many enemies' hitboxes will do damage before their weapon has even made contact”, matching a Korean review's “순간이동 판정에 맞았다” (hit by a teleport hitbox).
• Succeeding still pays too little to justify the risk, and new material confirms rather than extends it: a perfect parry does not stagger elites or bosses out of a combo, executions need three to six parries, and “idk who thought it was rewarding for your parry attack to do less damage than a regular attack”.
• No new review reports the parry windows changing in this window. The only timing-change claim is still the pre-existing one that release nerfed the demo's parry (56-helpful Chinese on shortened active frames and lengthened pre/post animation), and the single mention of a balance patch concerns enemy damage, not windows.

The theme grew 450 to 649 mentions, but every new sentence comes from a review with 33 or fewer helpful votes and restates the top-ranked accounts; only 40 of 1026 matching sentences were reviewed.

FindingPlayers are not attacking how the world looks — they attack how it reads and what it costs to cross: an open world where the terrain is hard to tell apart, a map tool that arrives late and does little, and beacons too sparse to absorb the deaths. The weight has moved onto the second half of the map, which recurs across languages as padding rather than design.

• 31% of the reviews raising this still recommend the game
• Simplified Chinese reviewers raise it 2.1× more often than average (50% of their complaints)
• English reviewers raise it 2.2× less often (11%)

• A correction to the baseline lead: the layout itself is now attacked, and no longer only in Chinese — new English and Russian reviews name sameness and illegibility directly, most precisely as “The level design has improved vastly, but my god, is it dreary, repetitive and an absolute disaster to navigate because everything looks the same”. That matches the long-standing Chinese charge that “地图的辨识度很低” with no connective structure, so it recurs in three languages rather than one.
• What the baseline got right and still holds: the praise is real and comes from people refusing to recommend — “Map design is good, exploration is rewarded without being tedious” and “level design and presentation are absolutely peak” are both from non-recommenders. The split is between how the world looks, which is praised, and how it reads to move through, which is not.
• Checkpoint placement and what it costs remains the one charge that recurs in every language, unchanged in substance: “too few checkpoints in open world” (Polish), “lack of beacons/checkpoints” (English), and a Russian review asking why players are still sent to a boss from a bonfire “через полкарты”. English adds the placement version — “quite often the hardest encounters on the level are put right before the checkpoint”.
• The endgame point should widen from the final area to the whole back half, and the word players reach for is padding: “前面8分后面0分”, map, dungeon and monster design “完全的敷衍，全是为了拖长游戏时长”, and Polish “Game designed just to waste player time”. A Russian review dates the break exactly — after the first three bosses “мы перепрыгиваем на вторую половину карты где обман раскрывается”.
• Exploration's reward yield recurs across Chinese, English and Russian but is contested rather than settled: “因为我也不知道地图探索给过什么很有价值的东西”, “you can just run over everything and you will mis nothing”, “Exploration kind of sucks - it's not very rewarding” — against a German and an English review calling the same rewards genuinely worth finding.
• Russian reviews carry the map-tooling specifics at a rate no other language matches, and they stay narrow and fixable: zone maps sit “в самом конце локации”, so you clear the zone blind first, and the warp gates want clearer drawing — “можно было лучше и внятнее рисовать проходы и ‘прыжки через врата’”. Russian is also the only language whose top recommender credits the map as the fix for the first game's confusion, listing as features the same self-marking dungeons the top Chinese review treats as a bad joke.

752 mentions is not 752 complaints — the keyword sweep catches praise, and several of the highest-read hits are compliments to level design inside reviews that do not recommend; with 480 of 752 in Simplified Chinese the count also overweights the legibility and padding charge relative to the checkpoint one, which is the only charge recurring in every language.

FindingThe complaint is not that beacons are too few in the abstract but that their density runs inverse to danger: clustered across empty ground, absent inside dungeons and across the open mob-dense stretches. A death there costs a resource-draining sprint back rather than a retry, and the fixes players name are small and specific.

• 27% of the reviews raising this still recommend the game
• Simplified Chinese reviewers raise it 1.5× more often than average (35% of their complaints)
• German reviewers raise it 2.0× less often (12%)

• The dungeon boundary is still the single most-repeated failure, and the new sentences restate it almost verbatim: “Also dungeons don't have checkpoints so you may need to run back to the dungeon as well as running through it”. The ask has not moved either — an Elden Ring-style stake at the entrance, or “为什么boss房门口不能有火” (why can't there be a fire at the boss room door).
• Scarcity belongs back in the diagnosis alongside placement, because the new batch says both plainly: “Checkpoints/bonfires are very scarce for how many and how far away they are from each other”. The inversion is what recurs across languages — two beacons side by side across “two completely empty fields with absolutely nothing happening”, while “有些地方八百里不见一个篝火” (in some places you go 800 li without seeing one) covers the open, mob-heavy ground.
• Boss doors are the exception, and the most-helpful new review is a recommender who states it: clearing an area yields several bonfires, “один обязательно рядом с основным боссом, и никакой беготни до него нет” (one always right by the main boss, with no running to it at all). So what the theme is really about is dungeon interiors and overworld traversal, not the retry loop at a boss.
• New in this batch: some of the reported scarcity is beacons not found rather than not built — they “just blend into the environment”, the layout “feels like a maze”, and a Russian reviewer asks why Elden Ring's bonfire markers weren't copied along with the rest of the open world. That makes map indicators a separate and cheaper fix than adding checkpoints.
• The runback is a resource tax, not just lost time, because nothing outside a beacon restores health: “No checkpoint, no healing refills”, “篝火数量稀缺，没有别的回血手段” (bonfires are scarce and there is no other way to heal), and a Russian count of three to eight flasks for a stretch holding ten thousand mobs and three dungeons. Currency compounds it — at 10% health mid-runback players “kiss your currency goodbye” rather than finish the walk.
• Emphasis still splits by language: Chinese (305, nearly half the theme) files it as traversal design, with mobs too strong to fight so “只能马拉松” (you can only marathon) between distant fires; Russian (70) counts it in flasks and lost farm; English (243) argues it against FromSoft precedent, now reaching past Elden Ring and DS2 to Demon's Souls as the previous worst.

Nineteen of the twenty new sentences come from non-recommenders and none raises a complaint that wasn't already on the list, so the +187 tracks the corpus growing rather than the problem widening.

FindingThe complaint is not that the game is hard but that encounters are decided by the one-shot: a hit players cannot read off the animation, that the shell sold to them as a safety net does not stop, and that costs a long run back and hours of banked currency. What the newest reviews add is the other half of the arithmetic — enemy health and damage rose while player damage and the payoff for defending well stayed flat, so a five-minute fight is erased by one unreadable hit.

• 24% of the reviews raising this still recommend the game — a low share, so this one tends to come with a thumbs-down
• Simplified Chinese reviewers raise it 1.8× more often than average (40% of their complaints)
• English reviewers raise it 1.5× less often (15%)
• German reviewers raise it 2.2× less often (10%)

• Bosses are praised in the same breath as they are condemned — “Bosses are really clever and unique, but its all one shot mechanics” — because the objection is damage standing in for design: “Just big sponge boss with a lot of one shot attacks that you have no time to react”. The same substitution has erased the tier line for ordinary enemies: the top new Russian review reports former mini-bosses as base mobs placed “в комплекте с парочкой бомжей, чтобы любая схватка превращалась в битву ваншотов”, while “their boss variant … would do significantly less damage”.
• Correction: the previous read that this is not a late-game spike does not hold on this evidence — the collapse is placed in the back half over and over (“the 2nd half of the game becomes a boring slog”, the Mammon region, the final area, and a new Chinese review that runs at “7/10” until the last two bosses deliver “纯粹的数值伤害，以及多填个0的血量” — purely numeric damage and health with an extra zero). What survives is that only the enemy's side ramps: “have my build maxed out, and the game STILL one-shots me with virtually any enemy in the final area”, while a boss's “очень быстрые атаки” one-shot “самую танковую оболочку в игре”.
• The top new Russian review measures both sides at once: a poke on a boss is “1/200 от хпбара”, a perfect parry “2/200”, so “с каждым приходится танцевать по пять минут, минимальная ошибка скорее всего отправит на респавн”. Chinese reviews had already itemised that flat player side — execution damage “就跟普通普攻敲几下差不多”, 5% per weapon upgrade — and still name the beta as the version where the same systems worked: “测试时期弹反的手感很棒…且处决条积累速度更快，敌人伤害也没那么高”.
• The narrowest fixable defect is unchanged: the one-shot voids the defensive system the game is sold on — “Being one shot by a normal attack through your block or harden ability is insane” — and the recurring instance is the unblockable grab, one of which “甚至能直接清空主角血条” (empties the protagonist's health bar outright). A Traditional Chinese reviewer puts it most precisely: in a game whose selling point is blocking to stagger and execute, the bosses answer with “超高速紅光指令投(無法格檔)”, very fast red-flash command grabs that cannot be blocked.
• What makes the damage read as broken rather than hard is that it does not match its animation and lands regardless of spacing — “lots of one shots from non-red attacks”, “split second attack animation and monstrous magnetization”, “零帧起手” (zero-frame startup), and now “actual one-shots from outside your camera”. New English and Chinese reports push this from tuning toward bug: “异次元的攻击判定” (otherworldly hitboxes), hitboxes lingering after an enemy interrupts its own animation, and an enemy that “slides towards you ACROSS THE WHOLE ARENA because you ran away from an attack”.
• The complaint is the price rather than the death: “the lack of beacons/checkpoints”, a long walk back, and a random mob one-shot on the way deleting the currency that took “условно часа полтора” to farm. So the tuning removes the fighting instead of hardening it — “slow, annoying chip damage or a sprint‑past‑everything run”, or as a new review has it, “the game transform into a no hit challenge”.

Most of the +132 is restatement of one-shot material already present in the highest-voted reviews, and 23% of the people filed here still recommend the game.

FindingPlayers endorse not being able to over-level; what they attack is that the upgrade budget does not cover what the full game asks you to fund — eight shells, sixteen weapons and Tar Stones — and that the spending is one-way. The new reviews quantify that rather than revise it: a 100% clear funds full upgrades for roughly three of eight shells, and the economy holds for about ten hours before breaking at the endgame where per-upgrade gains stay tiny and enemy damage spikes.

• 32% of the reviews raising this still recommend the game
• Simplified Chinese reviewers raise it 1.8× more often than average (40% of their complaints)
• English reviewers raise it 1.8× less often (12%)

• The design intent is still explicitly endorsed in the same words — “I like the idea that the game doesn't allow your character to completely overpower everything through grinding” — and a new English reviewer adds “I love the Shell system, and progression in general”. The new material sharpens what is actually attacked: a Chinese player who platinumed three playthroughs calls the shells and Tar Stones genuinely interesting, then “这游戏灾难的数值设计毁了这一切” — the game's disastrous stat design ruins all of it.
• The beta claim is unchanged and unchallenged by any new review — “they didn't change the resource allotment AT ALL for upgrade materials from the beta” — and a recommending Russian now puts a number on the shortfall: 100% clearing the entire game yields enough Glimmers to fully upgrade three of eight shells, and spent Glimmers cannot be returned. Two other Russians independently report finishing with two or three shells levelled and resources scraped “под 0”.
• Irreversibility remains the central mechanism, and the new reviews restate it as build lock-in rather than mere waste — “The progression and the way you level your shells and invest in equipment basically locks you into one build, and if you don't like it, too bad”, while the rare shell currency “discourages you from spreading it across different shells”. The layers that do refund produce friction, not anger: one Chinese reviewer's only ask is that SP be shared so swapping shells doesn't mean re-allocating every point.
• The endgame concentration holds, and the new evidence shows that the disagreement over whether upgrades are felt is a split by stage rather than by language: the same Russian review that calls shell levelling “весьма неплоха и имеет свой импакт” also reports the final boss taking almost no damage at maximum upgrade. The newest reviews land in the late half — “прокачка не ощущается ВООБЩЕ никак”, and “the second half feel much more like a chore than an adventure”.
• Tar Stone gating is no longer Chinese-only and its cause is now documented in two languages: the currency to take a stone or charm from 2 to 3 drops only from taking others from 1 to 2 — “拉出来再吃回去”, pull it out and feed it back in. That loop is why at most “around 24 tarstones” of 75 can ever reach level 3.
• Correction to the previous pass: beacon scarcity belongs to the early game, not to the endgame failure — a 61-hour reviewer states it “becomes less of an issue in the second half”. Losing dropped souls on a run-back still recurs — “you might as well kiss your currency goodbye”, now with reports of gloom spawning inside walls — but it is a first-ten-hours complaint, not part of the late-game breakdown.
• A recommending Russian names the fix and scopes it, which is the most useful new information: extend the existing weapon reset to shells and stones, since “сложности с прокачкой легко могут быть исправлены патчами” while the enemy behaviour he objects to cannot be.

The +105 mentions are corpus growth of the same complaint rather than a new one, and a minority remain keyword collisions — pure checkpoint-distance gripes, world-building and soundtrack criticism, and a dropped-currency visibility bug.

FindingAlmost the whole roster appears in the first few hours, so later areas can only escalate by re-statting and stacking the enemies you already know. The objection is not that the enemies are badly made — reviewers repeatedly say they look good — but that the game redeploys them as a substitute for new content, most visibly across the back third.

• 29% of the reviews raising this still recommend the game
• Simplified Chinese reviewers raise it 1.5× more often than average (19% of their complaints)
• Russian reviewers raise it 2.2× less often (6%)

• Front-loading remains the core claim and reviewers date it consistently: “you've seen about 70% of enemies after finishing the first 30% of the game”, “The enemies you discover in the first 2 hours of the game as the same enemies you will fight 20 hours in”, and now “you see 80% or more of all the enemy variety in the entire game in the first area”. The fractions differ; the seam they describe does not.
• The escalation levers are named directly — the same enemies “in different areas, with the primary differences being how much health they have, how much damage they deal, or how many of them are thrown at you”, restated almost verbatim in new reviews as “All of the same enemies were present as in the beginning just with more hp and damage.” What makes it fixable is that the re-statting is unsignposted, arriving “with no indication that they are different from the same exact enemy you ran into two areas ago”.
• New reviews sharpen where the reuse lands: enemies are not just repeated but placed without regard for setting, “copy pasted into zones they don't even fit thematically, just to fill space”. What reviewers ask for is per-zone identity rather than a bigger roster — “zones had more distinctive enemy types PER ZONE rather than reusing the same enemies over and over again”.
• Upward reuse into bosses is now the best-corroborated extension of the complaint and is still placed in the back third: “The majority of the bosses though are copy-pasted or just reskinned versions of elite and fodder enemies”, 把精英怪改个颜色变成boss (recolour an elite into a boss). One German reviewer reports the inverse — “In the later stages of the game, some bosses are reused as normal enemies”, with boss mechanics reused until they strip “other bosses of their identity and unique movesets” — which is one voice so far, not yet a pattern.
• The reuse is not confined to the roster, and the dungeons have a named template: “The environments as well get extremely repetitive, as they all largely look the same”, and 9 из 10 начинаются с лифта вниз (nine of ten start with an elevator down). The consequence stays disengagement rather than difficulty, now stated most precisely as “I didn't stop because the game became too difficult — I got bored because it became too easy and repetitive.”
• Enemy craft is still exempted from the complaint — “Aesthetically I quite enjoy the world and enemy designs in this game”, “Enemy design is okay (visually)” — and recommending reviewers name the identical defect and absorb it: “Enemy variety is somewhat limited, but the combat itself feels consistently satisfying, especially once your build starts coming together.”

Chinese reviews hold an unchanged ~45% share of mentions (179 of 401, previously 155 of 346) and still file this as budget padding with 换皮, 堆怪 and 数值 as one accusation — sharpest as a question no English review asks: 你这头盔枪兵从一开始出现到最后，除了数值提升有过换皮吗？ (has the helmeted spearman been reskinned even once, apart from stat increases?).

FindingThis is not vague disappointment: reviewers name a boundary. The release keeps the beta slice as its best-tuned section and scales enemy damage, health and density up after it without scaling checkpoints, soul income or the enemy roster to match — so the free build people bought on is still the best part of the game they paid for.

• 19% of the reviews raising this still recommend the game — a low share, so this one tends to come with a thumbs-down
• Russian reviewers raise it 1.8× less often (6%)

• The line is still dated precisely at the beta's edge, and that slice is still what players praise — new reviewers now name why: “It is by far the most polished and balanced section of the game very likely thanks to the Open Beta.” A new minority strand qualifies this without overturning it: three reviews say the warning signs were already visible, that even the demo area's enemies hit like trucks.
• The mechanical claim remains a squeeze from both sides, and the new material fills in the side the last pass had thinner — enemy damage, health and density rise past the beta boundary while checkpoints and soul income do not: “the lack of checkpoints in this game is probably the worst in the genre, even beating out Demons Souls”. A Chinese review completes the economy end: inflated mob stats make fighting pointless, so souls are near-unobtainable and levelling stalls exactly where incoming damage climbs.
• The content-overlap claim is unchanged in substance but now recurs in a third language and in quantified form — “There aren't many more enemy types in the core game compared to what was available in the Open Beta”, “in the beta you will see 70% of all the enemies in the game yet you will explore only 5% of the map”, and German “Gefühlt hat man 60% in der Demo schon gesehen”.
• “Patch #1 seems to have reverted many things to the Open Beta state of balance” still carries its original weight — it is the one line indicating the release genuinely differed rather than being misremembered — but nothing corroborates it as a fix. No newer review mentions improvement, and arrivals since repeat the same complaints in kind: “There was clearly a huge rework/overhaul of difficulty from the demo, and the game is so much worse for it.”
• The bait-and-switch framing now has a stated mechanism rather than only a tone: the polished opening outlasts the refund window — people who buy on a good first few hours “won't be able to refund the game after a few hours” — which is what the recurring Chinese 把玩家骗进来杀 (lure players in and kill them) points at. Their remedy is still a redirect rather than a refund: “the Demo holds about 90% of the joy and it's free”.
• Chinese still reports the theme at a materially harsher register than English — fraud outright (诈骗, demo骗) where English says bait-and-switch — but it is no longer the plurality language; English has pulled back ahead, 135 to 122. The separate allegation that the press build differed from the retail build down to the stone passives still appears only in Chinese, still rests on the same high-helpful review, and no new review repeats it: it remains a claim, not a verified fact.

The +79 mentions (229 to 308) roughly track the corpus's own ~40% growth, so this theme is not accelerating relative to the rest, and its recommend-share barely moved (19.7% to 19.2%).

FindingStill an environmental collision problem rather than a sensitivity one: the camera lodges in the geometry and bodies that the fight pushes players into, and lock-on fails at the same moment. What the new sentences sharpen is that second half — lock-on's failure is not only that it releases, but that it refuses to, holding the character to a target and taking movement with it.

• 27% of the reviews raising this still recommend the game

• The core diagnosis holds unchanged: the camera catches on geometry instead of letting it fade — “On ramps, the camera jerks up wards” and “Fighting by a wall is impossible, because the camera pushes in-ward and looks straight at the ground”. The shipped options still do not defuse it — the Action camera “got stuck on everything, regardless of if you turned the camera assist option on or off”.
• Players are still mostly not choosing the bad geometry, the fight puts them in it — “the camera systems are horrible if you get backed into a wall” — and in boss rooms the boss's own body is the occluder. One attack requires getting behind a boss that “LOVES to stick to walls of the arena”, which means “throwing your eyes out the window”; another new reviewer is trapped between a boss's corpse and a wall.
• Correction to the last pass: boss rooms did not replace cramped space as the location, and arena size does not predict the failure — “The final boss had one of the biggest arenas in the game and at the same time one of the worst cameras”, yet new recommending reviewers still name “tight spaces or multi-boss encounters”. The boundary those reviewers grant has not been retired.
• What is genuinely new is the opposite lock-on failure: it will not let go. “The lock on hinders movement imo and magnetizes you to the enemy you're locked on to as if attached by some invisible chain”, and the consequence recurs past that one reviewer — players dragged off ledges with “no way of stopping my characters movement as the lock on doesn't release”, a Russian whose camera never detaches from an enemy falling into a pit and pulls him in after it, a Chinese reviewer whose front-facing execution wrenches the view until he falls to his death.
• Both older lock-on failures persist unchanged — “the lock on will randomly disengage for no reason”, and it is “hard to swap targets if there's any form of terrain” — and they now sit next to animations that remove the view outright. Knockdowns and abilities “force an unhelpful camera angle, and not allow you to readjust it”, a Turkish reviewer calls locking the camera during the riposte animation idiotic, and after an animation you are “laterally locked and unable to effectively turn, move the camera or switch targets”.
• Chinese is now 111 of 315 but contributed no new sentences this pass, so its materially different half is unchanged: soft lock only, targets switching on a slight mouse movement, being locked to one of two caged enemies while the other connects. The ask stays as specific as before — a hard lock plus auto-nearest targeting for mouse and keyboard, where English reviews instead ask for customisable camera placement and for intervening geometry to fade.
• The lethal register recurs and now has a named mechanism rather than only hyperbole: “actual one-shots from outside your camera are a definite overkill”, a Ukrainian sees “вообще нічого” while the camera sits in collision and enemies spam, and the Russian final boss lands hits from beyond the screen edge.

Volume grew 16% with the recommend share flat at 27%, and most of the +43 is restatement or contentless intensity (“I HATE the camera”, “Camera lives its own life”, Russian sarcasm about a camera Miyazaki could only dream of), with four of the genuinely new sentences coming from one English reviewer.

FindingThe complaint is not that enemies track the player at all, but that tracking continues after the attack has committed — enemies rotate and slide mid-swing, so a correctly-timed dodge or step-behind still connects. Riding alongside it is a separable second issue: an aggro radius so large that enemies detect from far off and pursue without disengaging.

• 17% of the reviews raising this still recommend the game — a low share, so this one tends to come with a thumbs-down

• What recurs most precisely is correction during the animation, not before it: enemies turn and close after the swing starts, which is what invalidates the dodge. One player states the shared complaint exactly: “enemies MAGNETIZE to you so hard, that it straight up feels like teleporting, spinning around their own axis and still hitting you when you dodged right behind them in the last moment.”
• Distance correction is the same defect on the other axis and is named just as often — enemies “start their attack windup from afar then instantly snap to your face on swing”, “teleporting 10 feet with a swing”, lunges that “close the gap from the moon in an instant”. Russian reviews report the rotation form at a similar rate (“мобы могут завернуть 360 атаку” — mobs can bend a 360 attack around).
• Aggro range recurs as its own grievance and is fixable separately: enemies detect at long range and never break off, which is what makes the runback hostile rather than just long — “some enemies also have an absurd aggro range and will happily follow you all the way into the boss room.”
• Almost no one argues magnetism alone would sink the combat; it recurs as the multiplier on one-shot damage, unstaggerable trash and five-enemy packs. A Russian reviewer who still recommends the game marks the threshold: with crowds, magnetism and no stagger stacked, “бой перестаёт быть интересным челленджем и становится душным” — the fight stops being an interesting challenge and becomes suffocating.
• Chinese reviews report a form the other languages barely raise: the adhesion pulls the player too, with dodges snapping onto bear traps and mob shoves sticking you into one for an uninterruptible kill — 捕兽夹 (bear trap) plus 吸附 (adhesion) appears across three of the top-voted schinese quotes.
• A smaller English-only strand attributes it to input reading rather than hitbox tracking — “alot of enemies have insane input reading” — worth keeping distinct, since it names a different cause and a different fix.

229 mentions at 17% recommend, but several of the highest-voted mentions (347 and 83 helpful) come from reviews that still recommend the game and list magnetism as the thing to patch — this is a fix-this item, not only a refund reason.

FindingPlayers are not objecting to dying to traps as such; the recurring charge is that the bear trap is unreadable against the floor, resets itself, and catches the dodge roll, turning it into a stun that the surrounding enemy pack converts into an unavoidable death. The cost is amplified by respawn points sitting far from where the traps are.

• 19% of the reviews raising this still recommend the game — a low share, so this one tends to come with a thumbs-down
• Simplified Chinese reviewers raise it 2.1× more often than average (8% of their complaints)
• Russian reviewers raise it 1.6× less often (2%)
• English reviewers raise it 3.3× less often (1%)

• The object almost everyone names is the bear trap, not traps in general: it is invisible against the ground texture, it re-arms itself after being sprung, and the dodge roll snaps the player onto it. The most precise statement of it: “另外各个地图里还遍布不少防不胜防的捕兽夹，踩完之后夹子还会自动重置，等着下一次踩中。”
• What recurs is the trap as a stun inside a mob pack rather than a one-shot by itself — rooted, the player is finished off by enemies they could otherwise handle: “喜欢被小怪推然后吸附夹子的不解释秒杀小连招吗？” A Russian reviewer reduces it to the same thing: pinned against a wall, you can let go of the keyboard.
• Exploding pots and barrels are the second recurring object, and the complaint is their placement inside active fights plus a blast radius wider than its own visual effect: “爆発系の攻撃やトラップの範囲がエフェクト以上に広範囲で避けづらい。” Several describe dying to a vase they broke while swinging at an enemy, not to a hazard they walked into.
• Trap deaths are described as expensive rather than merely frequent, because reviewers pair them with checkpointing — “the respawn points are so far apart that you'll just be spending the first few hours clearing the same areas and running back after you randomly die to a vase blowing up.” One puts it as 就连老贼都知道要放复活点 (even FromSoftware knows to place a respawn point).
• The two recommending reviews do not dispute the placement: they note the traps also damage enemies and that the irritation is a first-encounter one (初见坛子和捕兽夹有点烦人). That aims the fixable part at making a trap legible before you step on it, not at deleting traps.
• Simplified Chinese is 76 of 123 mentions and carries nearly all the specific mechanical charges — auto-reset, dodge magnetism, trap floor indistinguishable from plain wood. English reviewers more often skip the mechanics and question the design outright: “why are the bear traps even in this game?”, calling them a pacing interruption.
• Several reviewers group bear traps with hidden pits, corner ambushes and unprotected ledges as one pattern: cheap punishment that rewards memorisation, not play — the Polish review calls them mechanics that only create the appearance of difficulty.

The count is modestly inflated: five of the forty quoted sentences come from a single Traditional Chinese review using 陷阱 to mean a trap build or trap boss (版本陷阱巨劍), not a level hazard.

FindingThe attack is not on difficulty but on the fact that the numbers, rather than the player's play, decide the outcome: one or two weapon/shell combinations trivialise the game while most of the rest are non-viable, and late-game enemies are handed damage instead of behaviour. Both the recommending and non-recommending camps ask for the same narrow remedy — a stat pass, not a redesign.

• 26% of the reviews raising this still recommend the game

• The single most repeated specific is that heavy weapons are not worth using — slower without compensating damage or stagger — and that play converges on one answer instead: “the heavy slow weapons suck, they don't stagger enemies at all, just go for either of the dual wield options + duality stone.” Two Chinese reviews name the same build independently (axe-dagger plus the Duality Stone), and an English recommender says “only the Axe is any good, the rest do pathetic damage/stun.”
• The second recurring specific is scaling: later areas add health and damage rather than new behaviour, with Chinese reviewers pointing at the post-Mushroom-Village and Mammon maps where enemy stats are “carelessly filled in” and “any monster can one-shot the player.” The most-helpful review states the fix precisely — “a substantial balance pass that brings enemy health down and makes later encounters more mechanically interesting rather than simply more durable” — and three further English reviews ask for a “balance pass” in those words.
• Reviewers contradict each other on whether the game is too hard or too easy — one calls it “about the easiest soulslike I've played,” others are one-shot by trash mobs — and that spread is itself the finding, since difficulty is set by build choice rather than by play. One recommender states both halves at once: “enemies have crazy attack patterns and a lot of damage, but you are also crazy op at the same time.”
• Progression is repeatedly reported not to register, which is why the imbalance reads as unfixable by the player: “Late game riposting even with characters with nodes dedicated to it do PATHETIC damage,” “most of the upgrades I find are borderline useless,” and in Spanish “no siento mejoras al mejorar equipo o armas.”
• Shell balance is the quieter half of the same complaint and is where both Korean reviews land — shells are “many in type and clear in concept, but the actual performance difference is large,” and weapon gaps are “large enough to change the genre” — echoed by “the character/Shell balance is completely off” and a Japanese reviewer asking whether Proxima is simply weak.
• The emphasis differs by language at a rate worth noting: the 11 Chinese reviews target 数值平衡, late-game stat tuning that “doesn't feel like it was tested,” while the 22 English ones concentrate on weapon-class balance, and the 3 Japanese ones are mildest, treating it as tuning still owed by patches rather than a broken product.

25.6% recommend is low, but several recommending reviews name balance as their main reservation, so the recommend split understates how widely the complaint is shared rather than overstating it.

FindingPlayers are not asking for a minimap — several explicitly refuse one. What recurs is that the game lets you mark a destination on the map and then gives you nothing in the world to follow, so the actual gameplay becomes hunting for the next beacon.

• 26% of the reviews raising this still recommend the game
• Simplified Chinese reviewers raise it 1.8× more often than average (2% of their complaints)

• The highest-voted reviews ask for an in-world pointer to markers that already exist, not a navigation system: the 407-helpful review suggests “adding a golden light pillar above all beacons or showing beacon icons on the map, so players have a clear direction when exploring,” and the 85-helpful Russian review asks for map markers “visible from afar like endless rays into the sky.”
• Players who do use the waypoint tool report it does not survive the trip into play — “You can put waypoint on it but it’s basically pointless because in game it does NOT show you a single indicator ANYWHERE to point you in the direction of your waypoint” — echoed by the German review’s custom waypoints that are “frequently not visible in the fog of war” and by a Chinese review calling the marker function buggy.
• The destination being hunted is almost always the next beacon or bonfire, which fuses this theme with checkpoint scarcity; the recurring Chinese comparison is Elden Ring, where “Sites of Grace have guidance telling you there is one nearby, and this game has none.”
• The map also fails to record where you have been, which recurs as the most concrete fixable ask: “lots of forks but I have no idea which ones I have already run and which I have not,” and in English as the request that dungeon icons “automatically switch to a ‘completed’ icon once the player has finished the dungeon.”
• Chinese reviews report this at a materially broader scope than English — 指引 there means objective clarity as much as wayfinding, as in three hours played while “not knowing what I am doing or what the dungeons and purification beacons are even for,” with several saying you need an online guide before you have a plan. English reviews stay narrower, on direction, compass and waypoints.
• A minority, all English, defend the absence, so a tracked quest marker is the wrong fix: “You aren’t being pushed along a linear track by some annoying glowing marker or a mini-map that spoils every surprise,” while another concedes “no minimap is fine but no compass... in a multilevelled world” is what makes it annoying.

The 38 mentions match on 指引/guidance, which in Chinese covers both wayfinding and not knowing the objective; roughly a third of the Chinese sentences are about what to do rather than where to go.

FindingThe complaint is not that the translation reads badly but that the localized builds ship half-finished: English strings are left in place across menus, item text and dialogue, and the gaps fall on text players need to make decisions with. It is read as a sign the game shipped unfinished rather than as a language-quality problem.

• 32% of the reviews raising this still recommend the game
• Simplified Chinese reviewers raise it 2.3× more often than average (2% of their complaints)

• What recurs in every language is incompleteness, not mistranslation: Chinese reviewers report "the localization is incomplete, there are always places still in English," Russian reports "every 5–10 lines are untranslated, it looks like a cracked torrent translation," and Japanese reports untranslated fragments at a steady frequency.
• The gaps land on consequence-bearing text, which is what turns an annoyance into a complaint: the most-helpful review (41) is about the Slayer's Mark description, "this should be translated as: once equipped, permanently disables achievements for this save," and the Japanese review notes untranslated strings appearing in progression-relevant places.
• This is a live target, not a fixed state — achievements shipped English-only and were patched within hours of Chinese negative reviews, Russian and Japanese reviewers credit fixes since the demo, yet one reviewer reports gaps appearing only after the first patch where there had been none.
• Simplified Chinese reports this at a materially higher rate (19 of 28) and in harsher terms than anyone else, describing both 简中 and 繁中 as "translated halfway" and the work as "amateur-outfit-level localization" that only moved after review-bombing.
• Outside Chinese the complaints are narrower and language-specific — French says the text is fine but the FR voice acting is not, Ukrainian reports the game launching in Russian despite a Ukrainian Steam setting — and several of those reviewers still recommend the game.
• In negative reviews localization is almost never the standalone grievance; it appears as one line in a bug list alongside crashes, clipping and vanishing weapons, cited as evidence of a rushed release.

A few of the 28 are keyword bleed — two English/German uses of "translate" in the figurative sense and two 乱码 reviews about corrupted save files rather than text — so the real localization count is slightly lower.