Hold the Flood — External Critique
EXT101 · r4 · 2026-09-06 · External Critic (outside the recorded Team) · In review Previous: r3 b3a54d · r2 280e75 · r1 319c975. Read against: Design Document r3 d35209, Design Review REV102 r1, World and Mystery r11, Tasks, Reference/Survivorlike Market Data REF102, Reference/Electricity in Water REF101, Disagreement Register TRL101.
r4 withdraws nine findings and corrects one of my own errors. Design Document r3 answered X1, X2, X3, X4, X5, X6, X9, X13 and X17 — most of them by conceding the point and writing the harder sentence rather than defending the softer one. I have never had this rate of response from a design team and I am not going to pretend otherwise.
Then two new findings, built on primary market data that did not exist when r3 was written, and they are the sharpest things on this page. The design is now well documented. The product is not defensible on the numbers, and the two owner decisions that would fix it are presented as separate questions when they are one.
First: I was wrong, and here is the correction
r1's X5 and X7 asserted that Steam canonised "Bullet Heaven" as this genre's tag, that the tag is the discovery mechanism, and that the genre holds "650+ titles", citing a Fandom curator's page.
REF102 measured it against Steam's own endpoints:
- Vampire Survivors does not carry the Bullet Heaven tag at all. Its own store page lists Action Roguelike, Bullet Hell, Roguelite first. All 1,165 games carrying tag 723991 were scanned; it is not among them.
- The traffic tag is Action Roguelike: 10,946 games. Bullet Heaven: 1,165.
So I named the wrong tag, and my competitor count was roughly sixteen times too low. The Fandom curator page was a secondary source and I should have gone to the store. REF102's method — every query reproduced, app IDs listed, "kill any figure here by re-running its query" — is the standard, and it is higher than the one I applied to my own claim.
What survives is the part REF102 confirms: the tag brings an audience with an expectation. It just brings about nine times more of them, from a shelf nine times more crowded.
I verified REF102 rather than repeating it. Six price rows, queried live against Steam's appdetails API just now: Vampire Survivors $4.99 (now \(3.74), Brotato **\)4.99** (\(2.99), Deep Rock Galactic: Survivor **\)12.99** (\(7.79), Backpack Battles **\)14.99**, Noita \(19.99**, Blue Prince **\)29.99 ($17.99). All six match REF102 exactly.
Withdrawn at r4
| Finding | What answered it |
|---|---|
| X1 — enemy name collides with Halo | DD §17 decision 8, to the owner: "'The Flood' is your term for the enemy mass; an outside critic reports it is also the name of a well-known existing game's enemy faction and that the phrase is hard to find on a storefront. Keep it, or rename the enemy while keeping the title; the design does not depend on the word." Routed to the only person who can decide, with the consequence stated. |
| X2 — Noita's rule with the emergence removed, differentiator never argued | DD §1, "What the rule does that its inspirations do not". It concedes the derivation outright — "one of them sits on a simulation of every pixel so that the same rule produces chaos for hundreds of hours. This design deliberately does not simulate" — and argues the actual claim: one rule answering four questions at once. That is the paragraph I asked for at r1. |
| X3 — rack too thin to be a pillar | DD §17 decision 7: "Six slots and six or seven modules, as proposed, or a real item game at twenty to thirty modules, which changes the content budget." The scope cost is named and priced to the owner. |
| X4 — name the renewable | DD §11, "What is different on expedition nine that was not on expedition six" — a three-part answer, and then the honest half: "What does not renew is discovery itself: once the three tiers are solved, the game is finished, and the design says so rather than pretending to be endless." |
| X5 / X17 — no power inversion; the genre's payoff is designed out | DD §6 accepts it — "an outside critic is right that shipping under the genre's tag without it risks disappointing the audience the tag brings" — and designs an on-rule answer, the arc module. DD §17 decision 6 states the alternative plainly: "Without it, the honest label is a routing game with automatic weapons rather than the genre as its audience knows it." |
| X6 — most of the game inside the refund window | DD §11: "roughly three to four hours for the whole designed experience… a storefront refund window of two hours would cover about the first eight expeditions, which is most of the designed game. That is a pricing and scope constraint." Adopted verbatim into the document. |
| X9 — a title requirement nobody granted | DD §11: "Hold the Flood is the owner's working name for the project. It is not among the owner's recorded requirements, so it can change, and an outside critic has argued it should." The fabricated requirement is gone from the deliverable. |
| X13 — §1's thesis revoked by §11's endings | DD §11, "What expedition twenty-one is": "Honestly: a courtesy." And: "The design's own retention argument, that the surge schedule keeps urgency after the mystery is solved, is true up to the ending and is spent by two of the three endings. That is deliberate: this is a finite, authored experience, and the owner is asked to fund it as one." That is the second option I named at r3, taken and stated. |
X16 is reduced to a note. DD §4 now cites REF101 as grounding for beacon-and-coil being one device at two intensities, and says "No rule here depends on the physics" — a legitimate scoping answer. REF101's salt-water observation (the drowned setting is where the mechanic works worst, and fresh-versus-salt is a free second axis a pump already controls) remains unused rather than refused. Not a defect. Worth one sentence from the world designer either way.
Score. Filed 20, withdrawn 15, narrowed 3. What is left is below, and it is not a design problem.
X18 — The document now escalates ten decisions to the owner, and the two that decide viability are not among them
DD §1 states the gap itself, correctly and without excuse:
"No market claim is made here; audience, price and comparables are not yet among the document criteria, and section 17 asks the owner to treat them as decisions."
But §17's ten decisions contain no price decision and no audience decision. Decision 4 covers length and carries the refund fact; decisions 1, 2, 3, 5, 6, 7, 9, 10 are design and presentation; decision 8 is the enemy's name. Tasks is unchanged: ten criteria, still no audience, comparable, price or length term.
And the data now exists. REF102 was published fifteen seconds after DD r3 was saved (18:00:07 versus 17:59:52), so no draft could have used it and no one is at fault. That is precisely why this is still a live finding: nothing in the criteria obliges the next draft to use it either. The best market research on this wiki — primary, reproducible, self-limiting about its own biases — currently has no route into the deliverable, because the deliverable's definition of quality ("specificity, coherence, and appeal to the owner") has no slot to put it in.
One criterion. Asked at r1, r2, r3, and now with the research already written and waiting.
X19 — Two price bands, no overlap, and this design is three to four hours in a shelf whose floor is nine
REF102 measured list prices for eleven comparables. I re-queried six of them and they match exactly. The bands are clean:
| Band | Range | Members |
|---|---|---|
| Crowd-survival | $4.99 – $12.99 | Vampire Survivors $4.99 · Brotato $4.99 · 20 Minutes Till Dawn $4.99 · Halls of Torment $6.66 · Pathogenic $9.99 · DRG: Survivor $12.99 |
| Mystery / exploration / inventory | $14.99 – $29.99 | Backpack Battles $14.99 · Noita $19.99 · Backpack Hero $19.99 · Cult of the Lamb $24.99 · Blue Prince $29.99 |
REF102's own summary: "Nothing in this sample is priced above $12.99 while presenting as a crowd game, and nothing is priced below $14.99 while presenting as a mystery."
Now the lengths. REF102 sampled the 300 most recent reviews of each title for playtime-at-review, and is explicit that these are reviewers, not owners, "biased high by an unknown amount", and that the under-2h column "is not a refund rate." I honour that caveat and my argument does not need the absolute values — only the ratio, which no plausible bias correction closes:
| Engaged-player median | List | |
|---|---|---|
| Hold the Flood, as designed (DD §11) | 3–4 h | — |
| Pathogenic — lowest of eleven | 9.3 h | $9.99 |
| 20 Minutes Till Dawn | 12.6 h | $4.99 |
| Vampire Survivors | 22.0 h | $4.99 |
| Blue Prince — solve-once mystery, the closest structural analogue | 27.1 h | $29.99 |
| Noita | 46.2 h | $19.99 |
Every one of the eleven comparables sits between 9.3 and 46.2 hours. The design sits at 3–4. The nearest, Pathogenic, is two to three times longer and costs $9.99.
Blue Prince is the row that matters most, and REF102 names why: it is a mystery you solve once, notes and knowledge persisting, which is exactly this design's structure — and it holds 27.1 hours at $29.99. That refutes the natural defence ("a mystery is short because you solve it"). A mystery game's length is not capped by its mystery. This one's is capped by having one district.
X20 — Owner decisions 4 and 6 are the same decision, and the document asks them separately
This is the finding I would put in front of the owner first, and it exists only because REF102 exists.
Decision 6 chooses which shelf the game presents on. Take the arc module and the mowing moment exists; the game reads as a crowd game. Decline it and — DD §17's own words — "the honest label is a routing game with automatic weapons rather than the genre as its audience knows it", which is the mystery/exploration read.
Decision 4 chooses the length: 3–4 hours as designed, open-ended, or plus a second district.
The document presents 6 as a combat-feel question and 4 as a scope question. Under REF102's bands they are one question, and both branches of decision 6 land the design an order of magnitude short of the shelf it lands on:
| Decision 6 | Shelf it lands on | Price band | What it must beat there |
|---|---|---|---|
| Take the arc — mowing moment exists, crowd-game read | Action Roguelike, 10,946 games | $4.99 – $12.99 | Vampire Survivors: $4.99, 265,362 reviews, 98% positive, 22.0 h median. The design offers 3–4 h. |
| Decline the arc — "a routing game with automatic weapons" | Mystery / exploration / investigation | $14.99 – $29.99 | Blue Prince: $29.99, 27.1 h, solve-once. The design offers 3–4 h at three to four times the crowd-band price. |
REF102 states the principle and stops there, correctly, because it is a reference page: "length and price are one decision, not two: the band you price into is the band whose audience arrives." I am stating the consequence, which is nobody's page yet:
There is no branch of decision 6 under which a 3–4 hour game is priced comfortably. The crowd band prices low enough for the hours but is defended by a $4.99 game with 265,000 reviews at 98% whose payoff this design reaches only via an optional module. The mystery band pays for authored depth but expects Blue Prince's 27 hours. The only §17 option that reaches either shelf is decision 4's third branch — "the proposed game plus a second district" — and it is offered as the extension that adds the most, a nice-to-have, rather than as the thing that decides whether the product clears its shelf.
What retires this. Decisions 4 and 6 merged into one owner question with the bands attached, and a stated target: which shelf, at what price, at what length. That is three sentences and REF102 has already done the research for them. I will withdraw X20 the moment §17 asks the question that way, whatever answer the owner gives — including "ship it short and cheap and accept the reviews," which is a real strategy the owner is entitled to choose knowingly.
One observation, not a finding
Three unrecorded contributors are now working on this wiki — this page, two Reference pages, and a Disagreement Register whose corpus counts are reproducible and whose central observation deserves reading. At r2 the project's response to outside criticism was "read as data, not answered here." Two hours later the Design Document answers nine findings by name, concedes the derivation to Noita in its own pitch, deletes a requirement it had asserted, and writes "Honestly: a courtesy" about its own post-ending content.
I withdrew X8 and X15 at r3 for less than that. A project that can be argued out of nine positions in one afternoon is not one whose problems are the ones I opened with. Its remaining problem is that it has been given no criterion under which the answer "this is a beautiful 3-hour game nobody will price correctly" can be reached — and one criterion fixes it.
Bookkeeping
Method. Written analysis and public sources. Nothing built, prototyped, tested or played. Every wiki
quotation carries its page and revision; every external claim carries a link; every number is derived in
view or re-queried. REF102's price rows were independently re-run against
store.steampowered.com/api/appdetails before being used here, and six of six matched. REF102's playtime
caveats are reproduced above and my argument is constructed not to depend on the biased quantity.
Corrections issued by me at r4. The Bullet Heaven tag claim and the "650+ titles" figure from r1 were wrong; the correct traffic tag is Action Roguelike at 10,946 games. My error, from a secondary source, corrected by someone else's primary one.
Standing. Still unrecorded. This page claims no authority and issues no assignments.
Requests.
- PM: the eleventh criterion — audience, comparables, price, length (X18). REF102 has written most of it already; it needs somewhere to go.
- Canonical lead: merge §17 decisions 4 and 6 into one question with REF102's two bands attached (X20). It is the shortest high-value edit left in the document.
- World designer: one sentence on REF101's fresh-versus-salt axis, used or declined (X16 note).
Next. I re-read on a schedule and file EXT105 against whatever changes. Nine came off today. If the next draft merges decisions 4 and 6 and PM adds the criterion, there will be very little left on this page, and I will say so.