The Deep Water Nobody Is Fishing
Roblox is paying a 42% premium for a category of game almost no one on the platform is building. This report maps five years of platform data, tears down the ten titles that defined them, and names six concepts a two-person AI-first studio can actually ship into the gap.
Five things the data says, and one of them is a window
Roblox in mid-2026 is a platform in the middle of a deliberate identity change. It is trading short-term growth for an older, richer audience — and it is currently overpaying, in cash and in distribution, for anyone willing to build for that audience. That overpayment is the opportunity.
The platform is enormous and its growth just broke
Roblox closed 2025 with $6.8B in bookings and 126.5M average daily users — a 55% bookings year. Then age-check enforcement throttled the sign-up funnel. Q1 2026 bookings still grew 43% to $1.73B, but management cut full-year bookings guidance from roughly 22–26% down to 8–12% and guided DAUs to decline sequentially in Q2. The engine is intact; the top of the funnel is not.
Roblox is buying adults, and creators are the currency
Among age-checked users, 27% are 18+. In the U.S., that cohort monetizes over 50% higher than under-18s and the 18–34 segment is growing 50%+ year over year. As of 8 June 2026 Roblox pays a 42% higher DevEx rate — 37.8% effective, versus 26.6% — on U.S. 18+ spend in qualifying "novel" games. Same Robux, materially more cash.
The algorithm that made brainrot just got rewritten
In June 2026 Roblox extended "Recommended For You" from a 7-day retention proxy to a 28-day window, split play-through into separate session-quality and spend signals, and added a hand-curated Standout Games shelf for novel titles. Thumbnail-bait now decays. Depth compounds. The 2025 playbook is being actively de-weighted.
Roblox has published its own shopping list
In its Incubator and Jumpstart materials Roblox names the gap in plain language: RPGs, strategy games, and shooters are "heavily underrepresented despite strong demand from older age groups." It wants genre mash-ups, distinct visual style, and mechanical depth. It is rare for a platform to tell you exactly what it will promote.
The cost floor for high-fidelity work has collapsed
Roblox Studio now runs as an MCP server — Claude Code, Cursor, and Codex can drive creation flows directly. Cube 3D/4D generates textured, interactive assets from prompts; SLIM and Texture Streaming deliver console-grade fidelity to phones. The art-and-engineering headcount that gated "premium" Roblox games two years ago is no longer the barrier it was.
For roughly the next four quarters, Roblox will pay above-market cash, hand-curated homepage placement, and algorithmic preference to small teams that ship deep, visually distinct games for players over 18 — precisely the profile a design-and-writing-led studio with an AI-first pipeline can hit and a clone farm cannot. Playology's two structural advantages, visual craft and narrative quality, map onto two of Roblox's three stated novel-game criteria.
Five years of Roblox, read as a business
Between 2021 and 2025 Roblox roughly tripled its audience and grew bookings 2.5×. Then, in the space of two quarters, it chose to slow itself down on purpose. Understanding why is the whole game.
+55% year over year
peak 152M in Q3
35B in Q4 alone
up 63% from $923M
cut from ~22–26%
Revenue and bookings, FY2021 – FY2026 guided
Audience and engagement, FY2021 – FY2025
Where the money actually is
Roblox's headline user count is a poor guide to its economics. Monetisation is concentrated in a single region by a factor of seven, and the platform's own strategy follows that concentration rather than the raw DAU map.
Average bookings per DAU by region, Q1 2026
A U.S. or Canadian daily user is worth $38.93 a quarter. An APAC user is worth $5.47. Roblox's fastest DAU growth is in Japan (+96%) and India (+84%) — markets that add engagement and cost, not bookings. This is exactly why the company is willing to sacrifice sign-ups to convert its most valuable region into a verified, older, higher-spending audience. The DevEx premium applies to U.S. 18+ spend specifically, which is the same insight expressed as a payout policy.
The three mechanisms that just changed
What Roblox turned on
- 18+ DevEx rate. From 8 June 2026, $0.0038 → $0.005396 per Robux on U.S. age-checked 18+ spend in eligible games. Requires R15. No application needed.
- Standout Games. A hand-curated homepage shelf for novel titles, plus genre sorts and "Because you played X" in test.
- 28-day recommendations. Day 1, day 2–7 and day 8–28 signals replace the old 7-day proxy; session quality and spend now score separately.
- Incubator & Jumpstart. Six-month milestone programme, up to 40 teams per cohort, with Roblox engineering access and audience support. 8,000+ applications to date.
- Studio as MCP server. Claude Code, Cursor and Codex can execute creation and playtest flows directly against a project.
What that costs you if you ignore it
- Clickbait decays. Developers reported CCU collapses within days of the June algorithm change on titles with no 28-day reason to return.
- The mid-tier is brutal. Effective creator share sits near 24.5% without the 18+ premium. The median experience on Roblox earns nothing.
- Concentration is extreme. Top 1,000 creators took roughly 87% of the $1.5B pool; the top 10 averaged $38.5M each.
- Under-13 content is a compliance surface. Age-gating, chat restriction and active litigation make the youngest cohort the most operationally expensive to serve.
- The window is a window. Elevated DevEx rates are a customer-acquisition cost for Roblox. They will normalise once supply of 18+ content arrives.
Ten titles that defined five years
These are the games that shaped Roblox between 2021 and 2026 — by lifetime scale, by revenue, or by proving something new was possible. Read them not as a leaderboard but as a set of solved problems.
| Title | Live | Genre | Core audience | Lifetime visits | Est. rev / mo | Monetisation spine | Build profile |
|---|---|---|---|---|---|---|---|
| Rivals | 2023 | Competitive FPS | 13–24, male-skewed, PC-led | 15.6B | $8–15M | Weapon skins, battle pass — pure cosmetics | Small team, sustained live ops |
| Blox Fruits | 2019 | Anime action RPG | 10–20, global, mobile-heavy | 62.0B | $8–14M | Gamepasses, seasonal fruit rotation, update cycles | Studio-scale, 6+ yr live ops |
| Brookhaven RP | 2020 | Social roleplay / life sim | 7–14, near-even gender split | 83.4B | $6–12M | Houses, vehicles, VIP — low ARPU, vast volume | Small core team, low content cost |
| Adopt Me! | 2019 | Pet collection & trading | 7–14, female-skewed | 43.8B | $5–10M | Limited eggs, trading economy, premium pets | Uplift Games — full studio, ~40 staff |
| Pet Simulator 99 | 2023 | Gacha idle simulator | 8–14, mobile-first | ~10B | $4–8M | Crate RNG, aggressive Robux sinks, gamepasses | BIG Games — established studio |
| Steal a Brainrot | 2025 | PvP collect / meme idle | 8–16, mobile-dominant | 69.8B | $3–7M | Gamepasses, rarity chase, live "admin" events | Small team — meme-native velocity |
| 99 Nights in the Forest | 2025 | Co-op survival horror | 10–22, session co-play | 27.2B | $3–6M est. | Cosmetic camp upgrades, timed boosts, no pay-to-win | 3 people · 3 months |
| Grow a Garden | 2025 | Cozy farming idle | 8–18, broadest gender split | 21.2B† | $12M in May 2025 | Seeds, pets, cosmetics, event drops | Teen solo origin |
| Dress to Impress | 2023 | Fashion / social competition | 10–19, strongly female | 10.6B | $2–5M | Cosmetics, seasonal passes, brand collabs | Solo creator origin, age 14 |
| Murder Mystery 2 | 2014 | Social deduction | 9–17, evergreen | 21.6B | $1–3M | Knife cosmetics, crates, trading economy | Solo creator origin |
† Statista snapshot, Jul 2025; substantially higher today. Revenue figures are third-party directional estimates — Roblox does not disclose per-game revenue. Build profiles reflect creator disclosures where public.
What the teardown actually tells you
Team size is not the constraint. Systems clarity is.
Four of the ten either began as solo projects or were built by teams you could fit in a car. 99 Nights in the Forest was made by three people in New Zealand in a three-month sprint and peaked at 14.2 million concurrent players — a number 2.7× their country's population, and enough to sell theatrical rights to 20th Century Studios. Dress to Impress was started by a 14-year-old. The barrier on Roblox has never been headcount; it is knowing precisely which loop you are building.
Cosmetics beat power, and the top earner is a shooter
The single highest-grossing experience on the platform is a Call of Duty-style FPS monetising through weapon skins and battle passes. Dress to Impress sells outfits. Brookhaven sells houses. Self-expression, not advantage, is the dominant purchase driver — the same shift that reshaped Fortnite and Valorant, now fully arrived on Roblox. Gacha still works but is carrying rising regulatory and reputational cost on a platform this young.
Engagement and revenue have decoupled
Brookhaven and Steal a Brainrot dominate total minutes played but earn well below what their engagement implies. Rivals converts a fraction of their playtime into more money. Volume is not the business; spend intent per hour is. A smaller audience with a genuine reason to buy outperforms a vast audience with none — and after the June algorithm change, it also ranks better.
The tech stack is nearly identical across all ten
Luau in Roblox Studio, server-authoritative gameplay, DataStore persistence, Rojo plus Git for anything with more than one contributor, Moon Animator for character work. Nobody in the top ten won on engine sophistication. They won on loop design, update cadence, and — increasingly — on being clippable. Which means the technical moat is available to anyone, and the creative moat is the only one that holds.
Every one of them was designed for the clip
The deer in 99 Nights. The admin-war events in Steal a Brainrot and Grow a Garden that drove Roblox to a 47.4M platform concurrency record. The Met Gala recreations in Dress to Impress. Games that produce a fifteen-second moment worth recording get free distribution on TikTok and YouTube; games that don't, buy it. Design the screenshot before you design the economy.
Nine of these ten target players under 18. Not one is an RPG, a strategy game, or a narrative-led experience for adults. The most successful 18+-leaning title on the platform is a shooter that monetises like a mainstream FPS — evidence that the older audience is present, spending, and served by almost nothing.
Where the water is empty
White space on Roblox is not "a genre nobody has made." It is the intersection of four things: real player demand, thin supply, alignment with the 28-day discovery model, and eligibility for the 18+ payout premium. Very few categories satisfy all four at once. Those that do are listed below.
| Category | Demand | Supply | 28-day retention fit | 18+ premium fit | Verdict |
|---|---|---|---|---|---|
| Competitive FPS | Very high | Thin but fortified | Strong | Yes | Contested — Rivals owns it |
| Anime action RPG | Very high | Saturated | Strong | Partial | Closed |
| Cozy / idle farming | High | Saturated | Weak post-June | No | Closed |
| Gacha simulator | High | Saturated | Weak post-June | No | Closed & regulatory risk |
| Social roleplay / life sim | Very high | Saturated | Moderate | No | Closed |
| Co-op survival horror | High | Moderate | Very strong | Yes | Open at the premium tier |
| Extraction / run-based risk | High (proven off-platform) | Near zero | Very strong | Yes | Open |
| Strategy & social deduction | High, named by Roblox | Thin | Strong | Yes | Open |
| Narrative-led RPG | High, named by Roblox | Near zero | Very strong | Yes | Open |
| Roguelite / deckbuilder | Moderate–high | Near zero | Very strong | Yes | Open |
| Fashion & creation with depth | Very high | Broad but shallow | Moderate | Yes, if aged up | Open above age 16 |
| Romance / relationship-driven | Unmeasured on-platform | Zero | Unproven | Yes | Untested — asymmetric |
Four openings, in order of confidence
The run-based risk loop has no Roblox equivalent
Extraction and "greed versus survival" run structures are among the most successful designs of the last five years off-platform — and they are essentially absent from Roblox. The loop is ideal here: sessions are short and mobile-friendly, tension is inherently social, every run generates a clip, and the metagame between runs is exactly the day 8–28 retention signal the new algorithm rewards. It is also cheap to build well, because run-based games reuse a small set of assets across enormous variation.
Nobody is writing for the people who grew up here
The 13–17 cohort is 38% of age-checked users and ages into 18+ every year. They have spent a decade on this platform and there is almost nothing on it written for an adult sensibility — no sustained narrative, no consequence, no tonal maturity. Roblox names RPGs as its single biggest gap and is now paying a premium to fill it. Writing quality is the least clonable asset in games, and it is the one thing a Chinese-studio content farm structurally cannot ship.
Strategy is the cheapest premium genre to build
Turn-based and auto-resolved combat needs no netcode heroics, no anti-cheat arms race, no animation budget, and no twitch-latency infrastructure. It is systems work — the exact thing an AI-assisted pipeline accelerates most and where design taste matters most. Roblox has explicitly named strategy as underrepresented. This is the highest quality-per-dollar category available on the platform right now.
The female 16+ audience has volume and no depth to spend on
Roughly 39–44% of Roblox users are female. Dress to Impress has 10.6 billion visits on a loop with no progression, no economy, and no narrative — a demand signal far larger than the product serving it. Players who started on Royale High at eleven are now nineteen with disposable income and nowhere to go. Any experience that gives this audience real craft systems, a player economy, or an actual story is competing against near-zero supply.
Roblox's novel-game criteria are genre, gameplay depth, and visual style. Playology's two structural strengths — visual design craft and writing quality — hit two of the three directly, and the AI-first pipeline (Claude Code against Studio's MCP server, Cube 3D/4D for assets, Suno and ElevenLabs for audio) closes most of the third. This is an unusually clean fit between what a platform is asking for and what this studio is already good at.
Six things worth building
Each concept below is scoped for a one-to-two-person AI-first team inside a $25–50K pre-launch envelope, targets a named white-space vector, and qualifies for the 18+ DevEx premium. They are ranked by expected value, not by ambition.
Four players crew a submersible working derelict rigs and wrecks on a drowned continental shelf. You descend, you strip what you can carry, and you surface before your air, your hull integrity, or the thing circling your lights runs out. Everything you extract funds the boat; everything you lose is gone. A rotating weekly wreck seed keeps the map fresh without a content treadmill.
Why it wins: darkness is the cheapest high-fidelity art direction that exists — volumetric fog, caustics and a single moving light source produce genuine "wait, that's Roblox?" fidelity at a fraction of the asset cost, and SLIM plus Texture Streaming carry it to phones. The greed-versus-survival decision generates a clip every single run. The boat upgrade tree and crew ranks are a purpose-built day 8–28 retention engine.
A persistent court intrigue RPG set in the Ashveil world. Players take a house allegiance, build standing through seasonal story chapters, and negotiate alliances and betrayals that resolve server-wide. Written narrative is the core loop, not the wrapper — chapters land weekly, and player choices at the house level change what the next chapter contains.
Why it wins: romantasy is among the fastest-growing categories in fiction, overwhelmingly female, and centred on exactly the 18–34 cohort Roblox is paying to acquire. On-platform supply is zero. It draws directly on existing Playology narrative assets, which collapses the most expensive part of the build, and the resulting moat is the one competitors cannot reverse-engineer: nobody clones prose.
You work the overnight shift at a coastal radio station. Callers phone in. Signals arrive that shouldn't. Your only tools are a mixing desk, a transmitter log, and the decision of what to put on air. Anomalies are procedurally assembled from a library of written and voiced fragments, so no two nights repeat.
Why it wins: one room, one desk, and audio carry the entire experience — this is the highest clip-per-dollar concept on the slate by a wide margin, and it is almost pure writing and sound design. Its strategic value is not revenue: it is a ten-week instrument for measuring Roblox discovery behaviour, establishing an age-checked 18+ player base, and proving the AI-first pipeline end to end before committing to the flagship.
A quarantined port town during a plague year. Ten players run competing trade houses; some are carrying the sickness and profit from spreading it. Each round is a public council vote layered over a private resource economy, so deduction has an actual price attached rather than being pure argument.
Why it wins: economics and voting logic, not combat networking — the entire game is systems and UI, which is where an AI-assisted pipeline is strongest and where the asset budget is smallest. Betrayal is the most reliably clippable moment in games. Roblox has named strategy as a priority gap and Murder Mystery 2 proves the deduction appetite has been evergreen on this platform for over a decade.
Draft a crew, position them on a grid, and watch the round resolve. Win and you climb a branching run; lose and you carry a meta-currency forward. Asynchronous ghost opponents mean matches never wait on a queue, and a weekly rotating rule mutator resets the metagame without new art.
Why it wins: zero real-time netcode risk, near-zero animation budget, and a metagame that is functionally bottomless — the single best structural fit for a 28-day retention signal on the entire slate. It also reuses design work already done on Playology's Spire & Saber deckbuilder concept, which materially shortens the build. Roblox has no credible entrant in this category.
Dress to Impress for people who have outgrown it. Source materials, unlock techniques, craft garments with real constraints, and run a boutique that other players actually buy from. Seasonal runway events are judged by the community, and the winning silhouettes enter the next season's material pool.
Why it wins: it takes the largest under-served demand signal on the platform and gives it progression, scarcity and a trading economy — the three things Adopt Me proved are worth billions of visits. The ceiling here is the highest on the slate. So is the operating burden: a player-run economy needs constant balance attention and active anti-exploit work, which is why it ranks sixth for a two-person team rather than first.
Do not open with the flagship. Ship The Last Broadcast in ten weeks for roughly $11K as a reconnaissance title: it costs little, it teaches you how Roblox discovery actually behaves under the new 28-day model, and it builds an age-checked 18+ audience you can retarget. Then commit the balance to SALVAGE with real data behind the decisions. Total pre-launch outlay across both sits near $45K, inside the envelope — and if The Last Broadcast performs, it funds a meaningful share of the flagship itself. If IP leverage matters more than genre fit, substitute The Ashveil Court at position one and the ladder logic holds unchanged.
Roadmap, money, and what could sink it
The window is time-boxed. Elevated DevEx rates are a customer-acquisition cost Roblox is paying to attract 18+ content, and they will normalise once that supply arrives. The plan below is built to be live inside it, not adjacent to it.
Validate and register
Submit to Roblox Jumpstart — it accepts applications on a rolling basis and requires no prior Roblox shipping history, which makes it the correct entry point. Stand up competitor telemetry against RoMonitor and Creator Analytics. Lock The Last Broadcast design document and its 28-day retention hypothesis before writing any Luau.
Build and ship the reconnaissance title
The Last Broadcast, end to end, on the AI-first stack. Non-negotiables: R15 avatars for DevEx eligibility, the Age Check API wired in from the first build, server-authoritative state, and Rojo with Git from commit one. Ship rough and ship on time — the objective is instrumentation, not polish.
Read the data, then decide
Four full 28-day cohorts. Measure day-1, day-2–7 and day-8–28 return rates against the published signal weights, test paid creator seeding against organic clip pickup, and establish real cost-per-installed-player. Flagship go/no-go gates here on evidence, not enthusiasm.
Build SALVAGE
Vertical slice by week 26; that slice is the Incubator application. Cube 4D for props, SLIM and Texture Streaming for the fidelity claim, Claude Code driving Studio's MCP server for systems work and automated playtest passes. Weekly wreck-seed generation must be running before launch, not after.
Soft launch and push for the shelf
Region-limited release, tune the extraction risk curve against live data, then drive for Standout Games placement. Creator seeding concentrates on the first seven days — titles that earn organic creator coverage in launch week are markedly more likely to hold above-average concurrency through month one.
Operate
Seasonal expedition passes, monthly wreck biomes, quarterly systems updates. Hold the cadence to what two people can sustain indefinitely — the most common failure mode on Roblox is not a bad launch, it is an update schedule that burns out the team that promised it.
Budget model
| Line | The Last Broadcast | SALVAGE | Note |
|---|---|---|---|
| AI tooling & subscriptions | $1,400 | $3,200 | Claude Code Max, Midjourney, Suno Pro, ElevenLabs |
| Contract art & mesh cleanup | $2,200 | $7,500 | Human pass over generated assets — the quality differentiator |
| Audio, VO & score | $1,500 | $2,800 | Higher share on Broadcast; audio is the product there |
| Plugins & licences | $300 | $600 | Moon Animator, building suites, analytics |
| QA & playtest incentives | $800 | $2,400 | Paid external cohorts; not friends and family |
| Launch UA & creator seeding | $3,500 | $12,000 | Roblox Ads Manager plus direct creator deals |
| Contingency (15%) | $1,300 | $5,500 | Assume one systems rewrite |
| Total pre-launch | $11,000 | $34,000 | Combined: $45,000 |
Recoupment scenarios
Modelled on developer cash after DevEx conversion, not gross player spend. Assumes a blended rate between the standard $0.0038 and the U.S. 18+ premium $0.005396 per Robux, benchmarked against public per-CCU earnings across the top ten.
| Scenario | Sustained avg CCU | Est. developer cash / month | Recoup $45K | Read |
|---|---|---|---|---|
| Fails to surface | < 300 | $1–3K | 15+ months | Kill and redeploy |
| Niche but alive | 800–1,500 | $6–14K | 4–7 months | Viable — iterate |
| Genuine hit | 4,000–8,000 | $32–75K | < 2 months | Scale live ops |
| Breakout | 25,000+ | $200K+ | Weeks | Hire, and hire fast |
Risk register
| Risk | Likelihood | Impact | Mitigation |
|---|---|---|---|
| DevEx premium normalises before launch | Medium | High | The ladder exists for this reason — be live and earning inside four months, not twelve |
| Discovery never surfaces the title | High | High | Jumpstart and Incubator placement, creator seeding in launch week, design explicitly for the clip |
| 18+ cohort smaller than headline | Medium | Medium | 27% is of age-checked users and only 45% had completed checks — treat all 18+ figures as directional |
| Platform DAU decline deepens | Medium | Medium | Weakness is top-of-funnel sign-ups, not retention or spend; existing-user monetisation held up through Q1 |
| Exploit load overwhelms a two-person team | Medium | High | Server authority from day one; avoid tradeable economies early — the reason Grand Atelier ranks last |
| Update-cadence expectations cause burnout | High | High | Choose procedurally reseeded content over hand-authored drops; commit publicly only to a cadence you can hold forever |
| Child-safety and regulatory exposure | Medium | High | 18+ positioning with the Age Check API is a compliance posture as well as a strategy; no gacha, no loot boxes |
| Generative-asset policy shifts | Low | Medium | Cube is first-party and Roblox-sanctioned; maintain a human-authored art bible so direction survives any tooling change |
Roblox has published its shopping list, raised its prices, rewritten its algorithm to favour depth, and handed every small studio an AI pipeline that makes depth affordable — and the correct response is not to build the biggest thing on the list, but to ship something small and instrumented within ten weeks so the flagship gets built on evidence.