Dusted

A1 · What exists today Built Open

Two unreleased prototypes and one set of competition rails. Verified 14 September 2026 on a clean checkout.^[P1]

Also measured and reported whether or not it flatters us. Pool fairness on the final table geometry: first breaker wins 50.74% over 14,400 matches, inside the acceptance band — and still no-go for cash until an unexplained between-batch dispersion is understood. A shot-selection solver held to human aim error is a coin flip against a strong scripted player; given perfect aim it wins 95%. Automated aim input is the threat that matters, and it is why cash limits will be governed by attestation, not by the shot clock.

Not done

• No proof it is fun. Passing tests establish nothing about enjoyment. An unfamiliar player choosing a second round is the gate, and we have not met it.
• No device. No iOS or Android build has run on a phone. Performance, thermals, touch ergonomics and the 60 fps target are unmeasured. Controls cannot be called polished.
• No real-money payments. No payment code paths exist; cash matches are refused outright. Deliberate.
• Three codebases, not one. The pool prototype and its server, the native racer, and the racer's ledger server are separate. Joining them is milestone one.
• No social layer. Friends, rivals, challenges, clubs, rematch and spectating are designed, not built. Matchmaking pairs the longest-waiting player with no rating.
• Authentication, transport and load are unexercised against real services: no identity provider configured, plain HTTP locally, no load test, no cost-per-match figure.
• Collusion and multi-accounting are undetectable before launch scale. Both need cross-match analysis over a population that does not exist yet.
• No licences, no channel approvals, no players, no revenue.

A1b · Defensibility stack and portfolio effect

A2 · Distribution & legal gates Policies read Sep 2026 No approvals held

Legal permission and platform permission are separate gates. No legal opinion is offered here: classification depends on consideration, prize value, the role of chance, player location, age, custody and payment flows, and “mostly skill” is not an exemption anywhere. Germany’s statutory test turns on whether the outcome depends predominantly on chance,^[S15] so adding randomness to frustrate bots moves that analysis the wrong way. The on-point precedent is Michigan: in October 2024 the Gaming Control Board issued a cease-and-desist finding that 21 Cash, Bingo Cash, Bubble Cash and Solitaire Cash constitute illegal gambling under state law.^[S34] A “skill game” label did not survive contact with a state regulator. Two risk vectors are easily conflated: the 2025–26 rulings that declared cash-entry products unlawful (Pennsylvania, Missouri) concern physical terminals, not mobile apps, so the read-across is doctrinal rather than binding, while the live exposure for a mobile operator has run through false advertising and consumer protection.^{[S29, S35]} Both are real; we do not plan for them the same way. Before any paid entry: jurisdiction-specific advice, operating permissions, payments acceptance, age and identity verification, geolocation, fund-handling and withdrawal terms, tax treatment, written Meta classification. Today every one of those is open.

And a category can be switched off. India’s Promotion and Regulation of Online Gaming Act 2025 banned real-money online games outright in August 2025; Dream11, MPL, Zupee and WinZO suspended paid contests within days, and the ban removed 95% of Dream Sports’ revenue.^[S32] That is not a prediction about Germany or the United States. It is the reason this round buys written permission in one market rather than scale across many.

Payments is a gate in its own right Checked 14 Sep 2026

The default payment rail refuses this business model, everywhere. Stripe's restricted-businesses list places under prohibited: "Games of skill including video game and mobile game tournaments or competitions, darts, card games and board games with a monetary or material prize." We checked three locale editions of that list — default, United Kingdom and Australia — and found no country carve-out in any of them.^[S38]

This holds regardless of whether the activity is lawful in a given market: the operator must be underwritten by a specialist gaming-payments processor rather than a general one — higher fees, real underwriting diligence, longer integration. We treat it as a funded work item in the permission gate. Whether PayPal and Adyen take the same position could not be verified. Open

Where we can operate — verified status only Verified Mostly unresearched

We are deliberately not publishing a green map. Of roughly twenty markets assessed, only a handful have been taken back to a primary statutory source; Ireland, Japan, South Korea, Singapore, Israel, Poland, Austria, Belgium, Spain, Portugal and most of Latin America remain genuinely unresearched. A market is listed above only where we have a basis. The plan funds one market properly rather than claiming many.

A3 · "Can you just copy these games?" Case law

The format, yes. The expression, no. US courts have said it plainly — "game mechanics and the rules are not entitled to protection" (Tetris Holding v. Xio, 2012).^[S25] Art, characters, names, levels and trade dress are protected: Xio lost because its clone was "all but indistinguishable", and the court noted it had "limitless options" to build the same-functioning game lawfully. The category runs on this — Stumble Guys cloned Fall Guys to 600m downloads.^[S17]

The real exposure is patents, not copyright. Nintendo and The Pokémon Company sued Palworld's developer in 2024 over three Japanese patents on a creature-catching mechanic — not over art. Damages sought: about $66,000.^[S26] Our rule: original art and audio from day one (already true — A7), freedom-to-operate review per format, and no mechanic that reads as someone's patent. Not legal advice.

The formats we have assessed

Every format below is short, thumb-playable, understood in ten seconds, and ends in a clean finishing order. We build originals of them, one at a time, on one competition layer. Orange marks the launch game and the next one.

Third-party titles are named for format identification only. Marks belong to their owners; figures are reported or third-party estimates describing those titles, not Dusted.^{[S6, S7, S17, S20–S24, S27]} Each Dusted game is an original build — no IP, characters, levels or branding used or implied.

Two of these already charge an entry fee. 8 Ball Pool and Golf Clash run coin-entry head-to-head matches every day, with a billion downloads and $789m of revenue between them — the only difference is what the winner is paid in.^{[S20, S21]} One already paid cash. HQ Trivia put real money on a casual game and reached 2.4 million concurrent players, then ceased operations in 2020; its prize pool was funded by investment rather than by a sustainable take on entries.^[S22]

A4 · The precedent, in full SEC filings

The public comparable’s revenue and payer history at its turning points.

Payer basis: FY2021–FY2024 figures are Q4 point-in-time; FY2025 and the 2026 quarters are as reported for the period. Marked where they differ rather than blended.^{[S8, S28]}

What broke, in one line: they rented the peak. Sales and marketing exceeded revenue every year from 2020 to 2022 and roughly half the 2021 spend was bonus cash. When Apple's ATT change raised the price of buying users and management cut the spending, the paying base fell from 610,000 to 110,000 over three years. Management's own quarterly language attributes the decline to their own marketing cuts, not to market conditions.

And the 2026 recovery is not what it appears. Revenue is growing while paying players keep falling. The growth comes from an AI ad-tech unit, from the Papaya judgment, and from charging a smaller base more per head. The company reorganised and renamed itself Firy Inc. (NYSE: FIRY) in June 2026 — confirm current corporate naming before circulating. Nobody has yet grown this business by growing the number of people who want to play.

A5 · Market and unit economics Model

US, iOS, permitted states. Pick a scenario; every figure below comes from one assumptions object.

Initial unit-economics hypothesis

At these assumptions Dusted needs about 5.7 paying months to recover a $60 acquisition cost; paying months are not calendar months. Retention and referral are the central economic questions.

One thing we will not do here is quote a market size for this category, because no honest one exists. The "$40bn global skill gaming market, 52% of it Asia-Pacific" figure that dominates search results comes from report mills, is not reproducible from any company's disclosures, and — decisively — no version of it we could find has been revised for India banning the largest real-money market in that region. Several still forecast Asia-Pacific as the fastest-growing region straight through the window in which it was outlawed. The only audited number in this entire category is the public comp's own: $541.9m of entry fees and $104.5m of revenue in FY2025.^{[S8, S32]} Everything below is built from stated assumptions instead, and should be argued with on those terms.

Context that is deliberately not in the front deck. Stake-adjacent culture is real — roughly two in five American men 18–49 have used crypto, 0DTE contracts are a record 66% of SPX options volume, and Kalshi, Polymarket, PrizePicks and Underdog were valued at about $22bn, $9bn, $2.5bn and $1.2bn in 2025–26.^{[S2, S3, S39]} None of them is a comparable for Dusted: liquidity, regulation, margins and user behaviour differ. They are evidence that people put money on outcomes for fun, nothing more.

Every row is rendered from one assumptions object; the front deck no longer carries the funnel, so the front deck and this table cannot drift apart. The reachable base is built bottom-up there: 260m US adults → 233m in states permitting cash entry → 212m smartphone → 123m iOS (Google Play bars the category) → 61.5m casual players. That last step is the one we have not sourced and it is labelled "to be verified" rather than asserted. Assumptions, not observed audiences or forecasts, and not a claim Dusted captures them. Before this is an investable SAM the first two rows need jurisdiction-specific adult reach and conversion evidence. Germany cannot be counted as reachable simply because Germans play games. We model at roughly a quarter of the only public comp's revenue per payer on purpose — Skillz runs ≈$320 of entries per payer per month at a 19.3% effective take, against our $150 at 10%.^[S8]

One structural point that is easy to miss and changes how every line above should be read. In this model, most entry volume is not new money. The public comp discloses that prior winnings made up more than 84% of total paid-entry fees — meaning roughly one dollar in six of gross entry volume is a fresh deposit, and the rest is the same balance circulating between players while the platform takes a cut of each lap.^[S37] That is good for gross volume and unforgiving about churn: when a cohort leaves, it takes the float with it. We therefore report net deposits alongside entry volume from the first pilot, and we do not treat entry volume as a revenue-quality measure.

Sensitivity: at an 8% fee and $9 of variable cost the conservative case pays back in 20 paying months — not a business. The variable line is the first thing to interrogate and we have no vendor quote for it. "Paying months" are not calendar months. No LTV is claimed. From the first pilot we report paid versus promotional entries, net deposits, fees, contribution and cohort retention separately. Our scale-proof target is sales & marketing under 35% of revenue, against the benchmark's 68%.

A6 · Integrity, liquidity and kill criteria Built Model

• Fair competition. Server-authoritative simulation; one hashed seed manifest producing an identical board for every entrant in a match; telemetry, replay review, anti-collusion, anti-multi-accounting. Bots in practice are labelled bots. Procedural variety is variety, not anti-cheat.
• Transparent money. Double-entry ledger; fee and split shown before entry; paid and promotional entries reported separately.
• Reliable resolution. Unresolvable matches void and refund exactly once — tested by killing the server mid-race. Late results are fenced and rejected.
• Never. No house bots in paid matches. No manipulated outcomes or personalised "luck". No paid performance advantages. The bar is no longer hypothetical and no longer an allegation. A jury found Papaya Gaming liable in April 2026 over undisclosed bots; trial evidence put 13 million bot accounts against roughly 11 million real players, and a court ordered $719m disgorged.^[S29] Separately, AviaGames faces an active class action alleging bot-seeded matches, with the Supreme Court having denied its arbitration petition in May 2026 — those remain allegations, not findings.^[S18] The distinction we draw is not that we would never do this. It is that our architecture is built so that a player, a regulator or a court can check. Published seed per event, replay available, human-only paid fields, and paid versus promotional entries reported separately from the first pilot.
• Liquidity — three tiers, not one. Eight players is a small number until you need eight specific ones. Pools fragment by region, stake, skill, device and mode. The answer is a ladder: continuous asynchronous entry as the floor, so a player at three in the morning still gets a game; hourly skill-bracketed live events as the spike; and one daily marquee with a concentrated pool. At 5,000 daily actives, 12% attendance on a single hourly event is roughly 600 simultaneous entrants — enough to fill seventy-five eight-player fields inside four or five genuine skill bands, where continuous matchmaking over the same population would be pairing a novice against a veteran or holding them in a queue. This is scheduled-tournament logic borrowed from poker, and the incumbent already ships all three tiers: asynchronous "play and compare", bracket events, and live events with a common seed per entry.^[S30] Hourly, not nightly. Tetris Primetime ran one cash tournament a night with over $1m a year in prizes and was discontinued; a single daily appointment gives a single conversion attempt a day.^[S31] The schedule is gated on measured actives — an hourly event with eight entrants is worse than no event at all. Measured at pool level, never as a platform average.
• Scheduled events make integrity provable, not just cheaper. One seed per event means every entrant ran an identical track, so outlier detection runs across a single large comparable field rather than across scattered pairings. After Papaya (the precedent table below) the ability to demonstrate that opponents were human and conditions identical — to a player, a regulator or a court — is a product requirement, not a compliance afterthought.

We stop if

• Approved channels only deliver payers at a cost needing permanent subsidy.
• Players who lose do not come back.
• Unfamiliar players will not re-run without a prize.
• Integrity cannot hold at pilot scale.
• No attractive territory clears legally.

A7 · Art provenance Built

All Dusted art is original and produced in-house: illustrated scenery, characters and the item atlas generated with recorded prompts and provenance; environment surfaces CC0 (Poly Haven) with verified hashes; Barlow Condensed under the SIL Open Font License; music and effects generated by our own tool. No third-party audio samples, no purchased or scraped game art, nothing from Nintendo or any other publisher. Third-party titles elsewhere in this document are named in type only, solely to identify a format, and the marks remain the property of their owners.

A8 · Sources Verified

Research cutoff 10 September 2026. Sources support the specific observations attributed to them — never the product, the modelled economics or the build claims. Platform policies must be re-checked before any campaign.

A9 · AI production claims Owner-supplied · to verify

The AI paragraph in section 08 rests on an argument about search cost that stands on its own: prototype cheaply, measure, kill, repeat. The specific vendor capabilities cited in support of it were supplied by the founder and have not been independently verified in this document. Each should be confirmed before it is repeated to an investor:

• Unity — project-aware AI agents inside the editor and an MCP server allowing external agents to inspect scenes, edit scripts and trigger editor actions. Confirm current availability and licence terms.
• AppLovin — automated generation of ad creative including video and playable formats. Confirm scope and whether it is generally available.
• Roblox — prompt-based creation tooling producing games at volume, cited as macro evidence that creation cost is falling. Confirm the figure and its basis.

The argument does not depend on any one of these being true. It depends on the direction of travel in engine tooling, asset generation and creative production, which is observable regardless of vendor. We would rather state that plainly than cite a capability we have not checked.

A10 · Hiring plan Model

One excellent launch game and a lean senior team. Headcount follows evidence; nothing below is committed beyond the first hire.

Legal and payments are bought, not hired: specialist gaming counsel and a gaming-payments processor, funded as a line of the round. What the founder de-risks: company and consumer product building, networked infrastructure and settlement, regulatory and payments research, fundraising and hiring. What the game director must de-risk: whether it is fun, depth, progression, live operations, game economy and multiplayer craft.

A11 · Integrity proofs Built

What the four promises in A13 look like in the prototype. Built claims are reproducible from the repository and describe payment-free software; amounts are from the model.

match  7f3a-0912
you    seed 9c1f7a…e07a
friend seed 9c1f7a…e07a
table  identical ✓

paid field
  you      verified human
  lina     verified human
practice
  opponent PRACTICE BOT ◀

replay 83d9…7630
turns re-derived    16 / 16
scores matched      16 / 16
VERDICT  reproducible ✓

entry   you        −$5.00
entry   friend     −$5.00
escrow  match      +$10.00
payout  winner     −$9.00
fee     dusted     −$1.00
escrow  balance     $0.00 ✓

0:41  server killed
0:41  match voided
0:41  refund you    +$5.00
0:41  refund friend +$5.00
0:42  refund again  REJECTED

Collusion and multi-accounting are open problems before launch scale; both need cross-match analysis over a population that does not exist yet. The Papaya and AviaGames records are in A6 and the sources.

A12 · Milestone gates Model Thresholds to set before pilot

What the round has to show, in order. Where a threshold is not yet founder-approved it says so rather than inventing one.

A13 · Trust promises and launch permissions No approvals held

A jury found Papaya liable in April 2026 over undisclosed bot opponents; a court ordered $719m disgorged.^[S29] Fairness has to be checkable by a player, a regulator or a court.

Dusted holds no licences or channel approvals today. Proofs in A11, markets in A2, integrity in A6.