The Worlds Report - Issue #7
- Roblox demand rose 2.0% week over week on a like-for-like basis, and eight of eleven genres grew. Last week it fell 7.7% and one genre grew. Simulator rose 6.6% and ended a seven-week slide in its share of demand. One week of contraction was not the start of a regime, and last week's issue treated it as one.
- We got the biggest finding in last week's issue wrong, and the correction is dated. We reported that Murder Mystery 2 grew 44.1M new visits having shipped nothing since July 23. Its content-update date now reads August 14, and July 23 turns out to be the launch of a summer event that runs to August 23. The game was mid-event the whole time. It is the biggest accelerator again this week at +56.7M.
- Fortnite's GTA-6 theme was not finished. We called it dead last week at zero islands in the top-15. This week GTA-VI fan islands went from 2.8% to 14.3% of Fortnite CCU, put four in the top-15, and account for 85.8% of the RPG/Adventure bucket. Rockstar announced on August 6 that its next showcase airs August 27, so this is event-driven and it has an expiry date on it.
- Survival/Co-op is vacant on Fortnite for an eighth straight week: 11.5% of Roblox new visits against 0.0% of Fortnite CCU. The evidence for what shape fills it got weaker, not stronger, and we say so below.
- Scorecard: 3 hit, 2 partial, 1 miss. Running 39 hit, 10 partial, 5 miss. Both partials are calls written as two conditions where the load-bearing half failed.
Running record: 39 hits · 10 partials · 5 misses
Last week's six calls, graded against this week's numbers. Three hit, two partial, one miss. Two calls were written as "A and B" and came back split; they are graded partial, not hit, and the failed half is named first in the evidence. A split condition is not a met threshold, and grading it as one would inflate the record.
- ◐ PARTIAL
Design for retention before monetization. Roblox changed what discovery rewards inside this window and said so publicly, so build the first session and the return visit before the earner. MED-HIGH
The vintage half failed. One of the top-5 accelerators predates 2020: Murder Mystery 2 (2014), against Steal a Brainrot (2025), Fisch (2024), Murderers VS Sheriffs DUELS (2023) and Candy & Chocolate (2026). The Horror half held at 16.8% like-for-like, past the 15% leg. The condition is not met, so this grades partial rather than hit. The mechanism behind the call is also falsified by its own example: Murder Mystery 2 was the old-title-winning-without-content evidence, and it turns out to have been running a dated seasonal event that started July 23 and ends August 23.
- ✓ HIT
Stop treating a content update as the growth lever on Roblox. This corrects our own call from last week. MED
Five of five. Animal Hospital -30.6M new visits (updated August 16), Mukbang Game -17.4M (August 17), Grow a Garden 2 -15.7M (August 15), My Avatar! -12.7M (August 15), Driving Empire -11.5M (August 14). Every one shipped inside the window and fell anyway. One qualifier we are stating rather than burying: updates also sat under four of the top-5 accelerators, so an update no longer predicts direction either way. It is a non-predictive lever rather than a dead one, which is a weaker and more accurate claim than the one we made.
- ✗ MISS
Do not build a 2025-vintage collect-sim (farm, pet or fishing) as an evergreen hedge. This is the advice we gave in last week's issue, and it did not survive the week. MED
Two of the three flipped to accelerators. Fisch gained 15.5M new visits (updated August 16) and Fish It! gained 0.9M (August 16). Only Grow a Garden 2 fell, at -15.7M. This is the second consecutive week we graded this cohort wrong, and in opposite directions: down with updates last week, up with updates this week. Fisch alone has run 23M, then 2.9M, then 18.4M across three windows. The honest description is a cohort that responds hard to content spend in both directions, and call 4 this week replaces the directional bet with that.
- ◐ PARTIAL
On Fortnite, drop the GTA-6 theme and build an escape room instead. They are cheap to build and they fit the platform's session length. MED
The GTA half failed hard. Four GTA-VI islands sit in the top-15: GRAND LOS CITY BOSS BATTLE (356 peak CCU), GDA VI MIAMI CARS (295), LOS GRAND CITY HORROR (259) and DRIVE CARS IN VA VI (204). The cohort went from 2.8% to 14.3% of Fortnite CCU. The escape-room half held at 9.8%, past the 8% leg. Condition not met. The driver was public before we published: Rockstar announced on August 6, inside last week's window, that its next GTA VI showcase airs August 27. We had that available and did not price it in.
- ✓ HIT
Take the Survival-to-Fortnite lane as a short co-op boss fight rather than a survival grind. The shape that cleared the top-15 this week is a bounded encounter. MED
Zero Survival/Co-op islands in the Fortnite top-15, an eighth straight week open. The genre is two islands and 5 CCU, which rounds to 0.0%, against 11.5% of Roblox new visits on the other side of the gap. The boss-fight half of the reasoning did not repeat: no co-op island holds the top-15 at 20 minutes per player or better this week, and last week's example is not in this week's figures at all. The metric passed and the shape evidence behind it is now one observation rather than a trend, so call 5 this week grades the genre bet and the shape claim separately.
- ✓ HIT
On VRChat, build the permanent destination rather than the update event. The crowd a content drop pulls in leaves again over the following weeks. MED
Both halves. FISH ǃ [PYRAMID] thinned again, three weeks running now since its content drop, while people carry on saving it. Popcorn Palace has shipped nothing across that same stretch and is still carrying the same crowd. This is the eighth consecutive week the destination-over-event read has held, and it held this week through a platform that drifted down as a whole, which is the harder version of the test.
Three things to know about this chart. It is Roblox only, because Fortnite has no usable cross-week genre comparison: island codes rotate through discovery and the two weeks share no islands at all, so any Fortnite week-over-week would be comparing two entirely different sets of islands. Genres under 1% share are left off so a small base cannot own the axis, which removes Roblox Other (0.5%) and Party (0.4%). And the two Roblox measures sit on different bases by design: share covers this week's full set of games, while the week-over-week compares only games present in both weeks. One entry needs a caveat of its own: Obby's +12.7% is one large returning course, which took 9.0M of that genre's 9.9M like-for-like growth.
Survival / Co-op PvE
11.5% of new visits
Roblox
0.0% of Fortnite CCU
Fortnite Creative
Eighth straight week. Fortnite Survival/Co-op is two islands and 5 CCU, with nothing in the top-15. Eight weeks is long enough that the vacancy is a property of the platform rather than an oversight, and the property is session length: Fortnite's median session ran 13.1 minutes this week, and a survival loop that needs half an hour to pay off has nowhere to land inside that.
What changed is the evidence for what shape fills it, and it changed against us. Last week we found a co-op boss fight holding 25.5 minutes per player and called that the version of the lane the platform would pay for. That island is not in this week's figures at all, and no co-op island holds the top-15 at 20 minutes per player or better. The turnover in island codes makes its disappearance unremarkable, but it also means the shape prescription now rests on one observation from one week. We are describing it that way rather than repeating it as though it accumulated evidence.
How a creator exploits it, with that caveat attached: build the deep version on Roblox, where a returning base compounds, and ship a bounded co-op encounter into the Fortnite lane. Keep it short enough to resolve inside the platform's session length and give it a win condition. Treat the specific boss-fight framing as a hypothesis under test rather than a finding, and this week's call 5 grades the genre bet rather than the shape.
A fourth relationship inverted this week, and it is not an opening. RPG/Adventure now runs 5.7% on Roblox against 14.1% on Fortnite, the first week in this report's history that the open-world bucket is heavier on Fortnite. It is entirely the GTA cohort, which is 85.8% of the Fortnite figure. That is a borrowed-IP wave with a scheduled expiry rather than a gap in supply. Build into it as a timing window if you want the August 27 spike, and price in that it ends.
One kind of event, three outcomes
Each of the three platforms met the same kind of input this week, a dated content event, and produced a different outcome. On Roblox (winner-take-all, where the top 10 hold 54.3% of new visits and the top title alone holds 12.4%) a seasonal event inside one incumbent moved the genre table. Murder Mystery 2's August 14 drop was worth 56.7M new visits, which on its own is larger than nine of the eleven genres' entire weekly totals. That is only possible on a curve where a handful of titles route most of the demand: the event does not lift a lane, it lifts the title, and the lane's number follows. On Fortnite (long tail, where the top island holds 3.2% of CCU, the top 10 hold 25.8%, and the median island peaked at 6 CCU on roughly 13.1 minute sessions) the dated event was external, a trailer announcement for a game that is not Fortnite. The long tail did what a long tail does: eighteen disposable islands appeared against the theme and collectively took 14.3% of CCU, with no single one exceeding 3%. Nobody built a franchise on it. On VRChat there was no event at all, and the platform drifted down slightly as a whole. The resting shape showed itself under that mild decline: the places built as places held level, while the place built around a content drop kept giving its crowd back. The creator-facing version is about what a dated event is worth on each curve. On Roblox an event is worth building toward, because the seasonal calendar is a demand schedule and an incumbent that runs one collects the whole genre's move. On Fortnite an event is worth building at, cheaply and immediately, then abandoning, because the codes rotate out regardless of how good the island was. On VRChat an event is the thing that costs you, since it rents a crowd you then have to keep renting.
Roblox
top-10 share 54.3% top-1 12.4%
Winner-take-all. The top 10 hold 54.3% of new visits and the top title alone holds 12.4%. Read these as levels rather than as a week-over-week move: the mix of games behind the comparison changed again this week, so the change in concentration is not a sound read.
Fortnite Creative
top-10 share 25.8% top-1 3.2%
Long tail. The median island peaked at 6 CCU on roughly 13.1 minute sessions. A fresh niche island is viable and disposable: eighteen islands appeared against this week's GTA theme and none of them exceeded 3% of CCU on its own.
VRChat, in brief
No VRChat platform event landed this week, so everything here is organic behaviour rather than a response to something the platform shipped. The platform as a whole gave back a little ground, and that changes what holding still means. A world that merely stayed level this week was doing better than the platform around it. Popcorn Palace, a chill-and-sleep hangout whose whole design is private rooms you bring your own people into, did exactly that, and it has now carried the same crowd for five weeks running without shipping anything at all. FISH ǃ [PYRAMID], the fishing game that grows by dropping content, kept handing back the crowd its update brought in, while people carry on saving it. That combination is what it looks like when a place turns into somewhere you intend to return to rather than somewhere you currently are.
The utilities split, and one open question closed
The avatar-search utilities split in a way worth watching. Prismic's Avatar Search, the place people pass through to find an avatar rather than to do anything, barely moved, and people keep saving it long past the point where a tool should have saturated. But avtr.zip, a newer world doing the same job with a hangout attached, is being saved nearly as hard while noticeably fewer people are actually standing in it. That is the same accumulate-intent-without-attendance shape we flagged on a race world a fortnight ago and got wrong, so it is worth watching and not worth calling. Avatar Testing Chamber, the mirror-and-rigging utility, quietly gained people in a week when the platform lost them. One open question from last week closed. We flagged The Spider-Lair: Evolved as an ambiguous case, saves running ahead of attendance with no way to tell a pre-load from a bookmark. It converted: people did show up. It has also already begun rolling over. So that pattern can be real and still be brief, which is the useful version of the answer, and it puts a rough three-week clock on how long the ambiguity lasts. What a builder should take from the week: when the platform drifts down, level is a win, and the worlds that stay level are rooms and tools rather than events. A hangout built around private instances and a utility people need often enough to keep saved both held their crowds through a soft week without spending anything on content. The world that had spent on content was still paying that crowd back.
Also pre-registered, ahead of the windows they grade
Three dated calls sit outside this week's six, registered here before the events they turn on. For the 2026-08-24 issue: Fortnite Chapter 7 Season 4 lands August 20 and ships a first-party visit-your-friends social space, and the call is that a first-party social island absorbs hangout demand rather than seeding a third-party lane. The metric is that Fortnite Roleplay/Life/Social stays below 1.5% of CCU and zero hangout or social islands enter the Fortnite top-15, in the window covering August 20 to 24. It sits at 1.4% this week, so the bar is live. For the 2026-08-31 issue: the Fortnite creature-event rental test, whose Discover placement ends August 25, graded on whether creature, pet and collection islands give back their event-window share afterwards. Not graded before then. For the 2026-09-07 issue: the Roblox discovery change reaches full rollout in late August or early September, and the call is that the genre effect strengthens at full rollout rather than fading as a test artifact, metered on like-for-like Simulator/Tycoon new visits falling while Horror/Escape holds at least 15%. This week's evidence runs against our own call, since Simulator rose 6.6% and demand recovered while the change sat in test.
Analyst notes
Roblox demand rose 2.0% this week on a like-for-like basis and eight of eleven genres grew. Seven days ago it fell 7.7% and exactly one genre grew, and we wrote an issue arguing that a discovery change was reshaping demand toward older, retention-heavy titles. This week’s numbers withdrew both pillars that argument stood on, so this issue opens with the corrections rather than burying them.
The first one is dated. We said Murder Mystery 2 gained 44.1M new visits having shipped nothing since July 23. Its content-update field now reads August 14, and the July date turns out to be a summer event launching, not a game going quiet. It was mid-event the whole time, and it is the biggest accelerator again this week. A content-update date tells you when something was last republished and nothing more than that.
The second is a calendar we did not check. We called Fortnite’s GTA-6 theme finished at zero islands in the top-15. It came back at 14.3% of CCU with four in the top-15, because Rockstar had announced an August 27 showcase on August 6, inside the window we were already looking at.
What survives both corrections is the update finding, now stated more carefully. Five of the top-5 Roblox decelerators shipped a content update inside the window and fell anyway, and so did four of the top-5 accelerators. An update does not predict direction in either direction, which is a weaker claim than the one we made last week and the one the data actually supports.
Each call below gets graded in next week's issue against the metric named on it.
-
On Fortnite, build a GTA-flavoured open-world driving island now and abandon it after August 27. The cohort went from 2.8% to 14.3% of CCU in one week on a trailer announcement, and the showcase itself is still ahead. MED-HIGH
Graded next week on: Graded 2026-08-31, in the issue covering August 27: GTA-VI-style islands hold at least 10% of Fortnite CCU and at least three sit in the Fortnite top-15.
-
Do not price the Roblox discovery change into a build decision until it shows up in two consecutive weeks. We built three calls on one week of contraction and two of them broke. MED-HIGH
Graded next week on: Like-for-like Roblox new visits do not fall below -2% week over week, and Simulator/Tycoon holds at least 24% of like-for-like share.
-
Build toward a dated seasonal event rather than toward a republish, and read a content-update date as "last republished" rather than as "has content." On a winner-take-all curve the seasonal calendar is the demand schedule. MED-HIGH
Graded next week on: Murder Mystery 2 remains a net accelerator in the window ending 2026-08-24, while its event is still running. Registered separately for the 2026-08-31 issue: it turns a net decelerator in the window covering August 24 to 31, once the event has ended.
-
Treat the collect-sim cohort as update-responsive and volatile, and stop calling its direction. We graded this cohort wrong two weeks running in opposite directions. MED
Graded next week on: At least two of Grow a Garden 2, Fish It! and Fisch that carry an in-window content-update date are net accelerators. If met, the cohort responds to content spend and is not structurally dead.
-
Survival-to-Fortnite is still the one real opening, but build it as a genre bet rather than as the boss-fight shape. The shape evidence did not survive the week. MED
Graded next week on: Still zero Survival/Co-op islands in the Fortnite top-15 by peak CCU. We will also report the minutes per player of any co-op island that does enter, so the shape claim accumulates evidence rather than being re-asserted.
-
On VRChat, build a room or a tool, and measure it against the platform rather than against zero. The platform drifted down this week, which means a world that merely stayed level was doing better than everything around it. MED
Graded next week on: A hangout or a utility world is still carrying its crowd next week while FISH ǃ [PYRAMID] keeps giving its own back.
- WARN The two Roblox columns sit on different bases, by design. Share covers this week's full set of games; week-over-week compares only games present in both weeks. Every Roblox direction claim in this issue rests on the like-for-like column.
- WARN We are not attributing a cause to the Roblox recovery. No dated public change landed inside this window. Roblox's week was regulatory rather than product, and the August 14 creator recap contains nothing on discovery, recommendations, monetization or payouts. The discovery change we wrote about last week is carried over rather than renewed, and it is still described as a test with an unfinished rollout. The recovery is reported here as an observed move with no cause attached.
- WARN We corrected our own headline finding from last week, and the correction is the week's load-bearing check. Murder Mystery 2's content-update date now reads August 14 where three consecutive weekly readings returned July 23. That single field overturned the reasoning behind last week's issue and re-graded call 1. A content-update date means last republished and never has content; July 23 was the launch of an event that runs to August 23.
- WARN No per-genre week-over-week is shown for Fortnite. The two weeks share no island codes at all, so any cross-week genre change would compare two entirely different sets of islands. Fortnite claims here rest on within-week share, top-15 composition, and per-island movement inside the week. Total Fortnite CCU is not comparable week over week for the same reason.
- INFO The GTA cohort figure is a composition comparison, not a same-island delta. The move from 2.8% to 14.3% compares the cohort's share of CCU across two different sets of islands, which is a within-week composition read on each side. Island first-seen dates carry no information on Fortnite, because effectively the whole set is new every week.
- INFO Roblox Obby's +12.7% is one title. A single large returning course took 9.0M of the genre's 9.9M like-for-like growth. The genre holds 2.2% of new visits and nothing else on the board has that shape.
- INFO Discount the small-base movers. Other at -46.1% and Party at -18.7% sit on 0.5% and 0.4% of new visits. Weight share and absolute size over week-over-week percentages on genres that small.
- INFO Spot-checks: three run, two clean, one that changed the issue. Murder Mystery 2's live player count read within 1.7% of ours, and its content-update field is what produced the correction above. A Fortnite island checked in the right band against a public third-party page, with a metric mismatch stated plainly: ours is a latest 24-hour peak and theirs is an all-time peak dated two days earlier, so ours reading lower is the expected relationship rather than an error. A VRChat world's favorites count reconciled to within 0.0005% of the platform's own public figure.
- INFO VRChat was checked on favorites and not on presence. Live presence is not verifiable without a logged-in session, so the VRChat read in this issue is qualitative, which is also where we keep it by policy.
- INFO A VRChat methodology error we caught and corrected during this run. An early pass bucketed the VRChat series by calendar week, which put the window's final day alone in a bucket of its own and made every world look sharply down. Recomputed on rolling seven-day windows, the real move is a mild platform-wide softening, and that is what the VRChat section above is written against.
- INFO Roblox's regulatory context, with a caveat attached. A US Senate child-safety investigation opened on August 12 with documents due August 31. It is not a genre mover and no demand movement this week should be attributed to it. It belongs here as a constraint on how optimistic anything about creator earnings should sound.
- INFO Nothing changed on the Fortnite supply side for a second consecutive window. The build held at v41.30 with no patch and no ecosystem release notes, so no Fortnite move this week has a supply-side cause. The first-party story moment on August 15 and the five-day event through August 19 are why Creative-side softness at the tail of the window reads as displacement. Chapter 7 Season 4 is dated August 20 and sits outside this window.
- INFO Method caveats. Fortnite island codes rotate through discovery, so per-island week-over-week is not possible. Fortnite genre placement follows an island's tags rather than its theme, which is why the GTA cohort sits in RPG/Adventure this week and sat in Simulator during the previous wave. VRChat reads stay qualitative and carry no per-world figures. Newly seen means a world first showed up in our figures, not that it launched.
Get the next issue by email
Every Tuesday, about 10:00 ET.
Free and weekly, with double opt-in. Unsubscribe any time.