Roblox Weekly: Steal An Egg grows 51.6% as the tracked cohort declines

A large winner inside a declining cohort

Steal An Egg rose from 1,053,720 to 1,597,564 average concurrent players between August 24–30 and August 31–September 6. Its increase of 543,843 average players was the largest absolute gain among qualifying games. The game accounted for 15.6% of this cohort’s summed weekly averages, up from 9.5% a week earlier.

That growth did not extend across most games in the sample. Of the 4,909 games with observations in every hour of both weeks, 1,021 increased and 3,888 declined. Their summed weekly averages fell from 11,077,099 to 10,256,602, a 7.4% decrease. Removing Steal An Egg from both weeks leaves a 13.6% decline across the remaining games.

The 20 largest games in the comparison

Ranked by average concurrent players during August 31–September 6, Steal An Egg led this cohort, followed by Brookhaven RP and Blox Fruits. Only 7 of these 20 games recorded a higher weekly average. A game can remain one of the largest while its audience declines from the preceding week.

CCU means concurrent users: the number playing at a sampled moment. Weekly averages are rounded to the nearest player for display; changes are calculated before rounding. Links open live game pages, whose current figures will differ from this snapshot.

The 20 biggest absolute gainers

Illegal Soccer climbed from 2,394 to 20,756 average concurrent players, adding 18,361. DOORS added 14,050 (+33.1%), while Anime Expeditions added 13,929 (+62.4%). These were substantial increases after accounting for their starting audiences.

Percentage changes answer a different question. Pull An Egg grew 1833.1%, from 418 to 8,073 average players. That is a much faster percentage increase than Steal An Egg’s, but its absolute increase was 7,655 rather than 543,843. We rank this table by the absolute increase and require a prior weekly average of at least 100.

This report measures changes. It does not attribute them to updates, promotion, recommendations, school calendars or other events that were not measured in this dataset.

The 10 biggest absolute declines

Adopt Me! recorded the largest absolute decline among qualifying games, down 71,886 average concurrent players (-26.8%). The next two largest declines were Steal a Brainrot, down 71,746, and +1 Speed Keyboard Escape, down 70,876.

A lower average during one complete week is not evidence that a game is permanently declining. These are two comparable periods, not a forecast.

How this report was calculated

1. Two complete UTC weeks. The current period runs from August 31, 2026 at 00:00 UTC up to, but not including, September 7 at 00:00. The preceding period is August 24 at 00:00 up to August 31 at 00:00. Each contains 168 hours, including the same weekdays.
2. A fixed, complete-hour cohort. We paired stored hourly records for the same game exactly seven days apart and required at least one recorded sample in every hour of both weeks: all 168 pairs and all seven days. This produces 4,909 games. A game did not have to be marked active at export to qualify. This rule favors games tracked throughout both weeks and excludes newly discovered or sparsely sampled titles. It is not a representative sample or census of all Roblox.
3. Equal weight for each hour. Within each hour, the tracker averages its recorded player-count samples and stores a rounded hourly mean. We take the arithmetic mean of those 168 stored means for each game and week. This gives each hour equal weight instead of allowing frequently sampled hours to dominate. The live site’s rolling, sample-weighted metrics can therefore differ.
4. Coverage does not mean continuous measurement. Sampling times and frequency vary. This cohort contains 4,238,723 underlying recorded samples in the current week and 5,203,332 in the previous week. Having a record in every hour does not guarantee identical sampling within those hours or that every short spike was captured.
5. Changes and selection. Absolute change is current average minus previous average. Percentage change is that difference divided by the previous average, multiplied by 100. The largest-games table sorts by current average. Gainers and decliners sort by absolute change and require a prior average of at least 100; ties use the universe ID. Cohort change compares the sums of game averages, rather than averaging their percentage changes.
6. A frozen export. The source was read from RoVitals’ stored hourly rollups at September 7, 2026 at 22:40 UTC. Names and genres reflect the export date; Roblox creators can change them later. Article numbers stay fixed while linked live pages continue to update.

Concurrent-player counts are not visits, unique people, hours played or revenue. We do not independently verify Roblox’s reported player counts or identify automated accounts. Averages and peaks can miss activity between observations. Read the site-wide collection methodology for source and tracker details.

Cite this report

When referencing the figures, include the dates, the 4,909-game cohort and a link to this report so readers can check the method. For a data correction or a question about the calculation, contact RoVitals.