Data & methodology
How BloxRadar knows what it knows
Real Roblox observations. Transparent market analysis. Evidence-based development decisions.
BloxRadar is an independent tool. It is not affiliated with, endorsed by, or operated by Roblox Corporation. Model versions in use: sat-1.0 (competition), fc-1.0 (benchmarks).
Sources
What we observe directly
Every observed statistic in the app comes from a public Roblox endpoint or from a stored reading of one. Nothing is invented, randomly generated or produced by an AI model.
- Experience metadatagames.roblox.com/v1/games — name, creator, visits, favourites, genre, created and updated dates.
- Live concurrent playersThe `playing` count in the same response, plus the public explore feed used for discovery.
- Votesgames.roblox.com/v1/games/votes — up and down votes, turned into an approval ratio.
- Place → universeapis.roblox.com/universes/v1/places/{id}/universe — resolves a game URL or place id.
- Thumbnailsthumbnails.roblox.com — decorative images only, never a data source.
Calculated
Style classification
Roblox genres are broad. We map each experience into a BloxRadar style bucket using its Roblox genre label plus keywords in its title. That makes the classification inferred, not official, and hybrid experiences can sit near more than one style.
Totals are summed over unique universe ids, so an experience that appears in several charts is never counted twice.
Calculated
Coverage limits
Charts describe the tracked sample, not the whole Roblox platform. Each panel shows how many experiences were analysed and when the latest sample was taken.
A sample is the trending and searchable experiences we can observe at that moment. It is large enough to compare styles against each other, and not a census of Roblox.
sat-1.0
How saturation is scored
Roblox publishes no saturation metric, so this is BloxRadar's own index. It combines four normalised components of the observed sample:
- breadth30%
- concentration30%
- volume25%
- Active competitor share15%
Breadth = tracked experiences / 30. Concentration = the leader's share of the style's observed CCU. Volume = log-scaled aggregate CCU against 250,000. Active share = the proportion of tracked experiences holding at least 100 concurrent players.
0–25
Low
Thin competition relative to the audience size.
26–50
Moderate
Contested, but a focused entry can still break out.
51–75
High
Established games own most of the audience.
76–100
Very high
Enter only with a genuinely different hook.
Fewer than 8 observed experiences in a style and we publish no score at all. A popular style is never treated as saturated on popularity alone: demand is weighed against how many experiences actually serve it.
fc-1.0
Comparable benchmarks, not predictions
Current CCU is observed. Future CCU for a game that does not exist yet is not available from Roblox, so we never claim to know it.
A brief shows the observed CCU of live comparables in the same style (25th percentile, median, 75th percentile) multiplied by an explicit entry-capture assumption of 25% / 50% / 100% of a typical comparable. Those are BloxRadar model estimates, not Roblox measurements.
- · A new experience competes for a fraction of what a typical live comparable already holds.
- · Comparables are the live experiences BloxRadar observed in the same style bucket during the latest sample.
- · No marketing spend, launch timing or retention difference is modelled.
Below 5 comparables, or fewer than 3 with real activity, the card says "Not enough verified market data to produce a reliable projection" instead of filling the space with a number. No confidence level is attached until the model has been calibrated against held-out launches.
History
Sampling and what a window means
Observations are stored with a UTC timestamp, one per experience per 15-minute window. A growth window is only shown when the stored series actually spans it — two days of collection never produces a "30-day growth" figure.
Until a window is fully covered, it reads "Collecting history" and shows the actual collection start date. Style roll-ups are stored hourly, so growth charts appear once enough hours have accumulated.
Reliability
When Roblox does not answer
Requests are batched, cached briefly and retried with backoff. If the upstream API still fails, the panel shows the last successful reading, marks it stale and keeps the stored history intact.
We never replace a failed response with generated numbers. A degraded sample is labelled as degraded, and anything that depends on it (competition scores, benchmarks) is suppressed rather than approximated.