Growth and momentum

How growth over 1 hour, 6 hours, 24 hours and 7 days, daily growth, the breakout score, velocity, acceleration, rising hours and volatility are computed.

Metrics & data / Metrics and data

Growth and the breakout signals compare a game with itself: its players now against its players earlier. The worker recomputes all of them every 5 minutes from stored snapshots, for every game read in the last 2 days that Roblox still serves (universe_metrics). Nothing here asks Roblox anything.

Momentum numbers at a glance

growth1h, growth6h, growth24h, growth7dratio (0.5 is +50%)

Window: the latest snapshot against one 0.9 to 1.5 windows earlier. Source: snapshots. Null: no snapshot in that range, or 0 players then.

growthDailyratio

Window: the last 24 hours' mean against the 24 hours before. Null: under about two days of history.

breakoutscore, 0 to 100

Window: the last complete hours. Source: hourly means. Null: under 100 players now, under 20 three hours back, or too few complete hours.

velocityplayers an hour

Window: the last 3 complete hours. Source: hourly means. Null: a missing hour.

accelerationplayers an hour

Window: the last 3 hours against the 3 before. Source: hourly means. Null: a missing hour.

rising hourshours, 0 to 6

Window: the last 6 complete hours. Null: unless all 7 compared hours have readings.

volatilityratio

Window: the last 24 hours. Null: fewer than 12 hours with readings.

How the signals are recomputed. Every 5 minutes the worker's metrics-tick recomputes every game in one statement. From the snapshots it picks each window's comparison snapshot and computes growth, the 24-hour averages and daily growth. From the hourly means it computes velocity, acceleration, rising hours and volatility, and from those the breakout score. Everything is written to universe_metrics, which the Radar, Explore and each game's page read.

How growth is measured

Growth is the change in players between the latest snapshot and one taken a fixed time earlier, as a share of the earlier count:

text
growth = (players now - players then) / players then

It is a ratio: 0.5 is shown as +50%, -0.2 as -20%. It is empty when there is no earlier snapshot in the accepted range (below), or when the earlier count was 0.

Which earlier reading a growth figure is compared with

Snapshots are not evenly spaced, so "one hour ago" is a range: the newest snapshot between 0.9 and 1.5 windows before the latest one.

WindowThe earlier snapshot must be
1h54 to 90 minutes before
6h5.4 to 9 hours before
24h21.6 to 36 hours before
7d6.3 to 10.5 days before

If no snapshot falls in that range the figure is empty, never a guess: a comparison across an unknown gap would report growth that is not there. After a collector outage, growth across the gap is empty by itself.

Growth in 1 hour

Players now against the snapshot about an hour earlier (growth1h). Shown as "in 1 hour" under Players now on a game's page and as 1h in Explore. A game read only hourly or daily may have no reading in the 54 to 90 minute range, so its 1-hour growth is often empty.

Player counts follow the time of day, so 1-hour and 6-hour growth are partly just the afternoon or the evening.

Growth in 6 hours

Players now against the snapshot about 6 hours earlier (growth6h). The Radar names it as a reason ("Up 40% in 6 hours") when it is 25% or more.

Growth in 24 hours (against yesterday)

Players now against the snapshot about 24 hours earlier (growth24h): Against yesterday on a game's page ("same time of day") and 24h in Explore. Comparing the same time of day removes the daily rhythm. It needs a day of history. The Market overview's Games rising today list is the games with the highest 24-hour growth among those with at least 100 players.

Growth in 7 days (against last week)

Players now against the snapshot about a week earlier (growth7d): Against last week on a game's page and 7d in Explore. It needs a week of history and removes both the daily and the weekly rhythm.

Daily growth

The mean players of the last 24 hours against the mean of the 24 hours before (growthDaily), so a single busy or quiet moment counts for little. It needs about two days of history: the first snapshot at least 44 hours old and at least 3 snapshots in each day; otherwise it is empty. Explore sorts by it as growth_daily.

Why small games with few players show huge growth

Growth is relative. A game going from 3 players to 9 has grown 200%; one going from 30,000 to 33,000 has grown 10%. Small games swing by large percentages from a handful of players, which is noise.

So the panel puts a floor on size wherever it ranks by growth:

  • Explore: choosing a growth column (1h, 24h, 7d, daily) to sort by adds at least 100 players (minCcu=100 in the address) unless you set your own minimum. Clear or change it in the filters to see smaller games.

  • Games rising today on the Market overview lists only games with at least 100 players.

  • The breakout score measures growth against at least 50 players and does not score games under 100 players (below).

The breakout score

One number from 0 to 100 for how much a game is breaking out right now, recomputed every 5 minutes from its hourly means. The Radar ranks by it. It adds four parts, then applies two penalties:

PartUp toWhat earns it
Momentum60Players gained over the last 3 complete hours, against the game's size; full at doubling
Acceleration15The gain speeding up against the 3 hours before; full when it gains as much again
Sustained15How many of the last 6 hours ended above the hour before (2.5 points each)
Youth10A new game: 10 when created today, falling to 0 at 90 days old
  • Swinging a lot costs up to half: when volatility is above 0.5, the score is multiplied by 1 minus the excess, down to 0.5.

  • Recovering, not breaking out: a game under half its observed peak keeps 70% of its score.

The size momentum and acceleration are measured against is the hourly mean 3 hours back, or 50 players if that is smaller, so a game going from 5 to 50 players does not score as a 900% rise. The result is rounded and capped at 100. The rule's reasons are in ADR-0011.

The whole rule, as the worker computes it (apps/worker/src/modules/metrics/domain/breakout.ts, and the same in SQL):

text
if velocity or the 3-hours-back mean is unknown:        empty
if players now < 100 or the 3-hours-back mean < 20:     empty
if velocity <= 0:                                       0

base         = max(hourly mean 3 hours back, 50)
momentum     = 60 x clamp(velocity x 3 / base, 0, 1)
acceleration = 15 x clamp(acceleration x 3 / base, 0, 1)
sustained    = 15 x rising_hours / 6
youth        = 10 x clamp(1 - age_days / 90, 0, 1)
score        = momentum + acceleration + sustained + youth

if volatility > 0.5:   score = score x (1 - min(volatility - 0.5, 0.5))
if peak_ratio < 0.5:   score = score x 0.7
breakout     = round(clamp(score, 0, 100))

A part whose input is unknown (acceleration, rising hours, age) adds nothing rather than emptying the score.

When a game has no breakout score

The score is empty ("Not scored yet", or "Needs eight hours of history" on a studio hub) when:

  • the game has fewer than 100 players now, or

  • its hourly mean 3 hours back is under 20 players, or

  • there are not enough complete hours to compare (it needs the last hour and the hour 3 hours before it; a newly tracked game needs several hours).

A score of 0 is different: the game has enough history and is not gaining players. The Radar lists only games scoring above 0, and the number of those is "scored".

Velocity

Players gained per hour over the last 3 complete hours: the last complete hour's mean minus the mean 3 hours before it, divided by 3.

text
velocity = (mean of the last complete hour - mean 3 hours before it) / 3

Shown as Velocity under a game's Signals, "+120 players an hour, over 3 hours". Negative means losing players. The Radar says "Gaining 120 players an hour" when it is positive. Explore and the Radar sort by it to break ties.

Acceleration

How the velocity changed against the 3 hours before: the gain over the last 3 hours minus the gain over the 3 hours before those, divided by 3.

text
acceleration = velocity now - velocity over the 3 hours before

Positive means the rise is speeding up ("Speeding up" on the Radar, and the Speeding up list holds the games with the largest positive acceleration).

Rising hours

Of the last 6 complete hours, how many ended with a higher mean than the hour before, from 0 to 6. The Radar names it when it is 4 or more ("Rose in 5 of the last 6 hours"). It is empty unless all 7 hours it compares have readings.

Volatility

How much the game's hourly means swung over the last 24 hours: their standard deviation divided by their average (the coefficient of variation).

text
volatility = standard deviation(hourly means, last 24 h) / mean(hourly means, last 24 h)

0.2 is shown as 20%. It needs at least 12 hours with readings. Above 0.5 the Radar says "Swings a lot from hour to hour" and the breakout score starts to lose points.

The radar lists

The Radar's four lists all hold games scoring above 0, from the same scores:

  • Breaking out: every scored game, best score first, then velocity.

  • Speeding up: scored games with positive acceleration, largest first (10 games).

  • New and rising: scored games created on Roblox in the last 30 days (10 games).

  • At an all-time high: scored games at 95% or more of their observed peak (10 games).

"Scores refresh every few minutes" means every 5. See Radar and Discovery for the page itself.

Related: Players and CCU, Tiers and genres, Glossary.

Updated