About the popularity index

Popularity is a relative signal for discovering browser games. It is not a player count, a revenue estimate, a quality rating or a prediction.

Poki and CrazyGames: voting signals

For each UTC date, we compare a game’s recorded cumulative votes with an observation around 30 days earlier. We divide the change by the actual number of elapsed days to obtain a recent daily voting rate.

• The end observation must be on or before the chart date and no more than two days old.
• The baseline observation must be on or before the date 30 days earlier and no more than two days older than that target.
• Future observations are never used for a historical chart point.
• A decrease between any recorded counters in that window is treated as a reset: the index is unavailable for that window.

Why a 0–100 scale?

We rank the eligible daily voting rates against other eligible games on the same portal, on the same date. Poki and CrazyGames are separate comparison groups. A value near 100 indicates faster recent vote growth than most eligible peers; it does not mean 100 players or universal popularity.

Tied rates share the midpoint of their ranks. With N eligible games, the percentile is 100 × (lower rates + 0.5 × (equal rates − 1)) / (N − 1). At least two eligible games are required. If the entire cohort has zero growth, we display “No recent voting activity” rather than a misleading middle score.

History is recalculated, not renamed

Every historical point uses its own rolling 30-day voting rate and that date’s same-portal comparison group. Changing the chart’s 7-day, 30-day, 90-day or one-year view changes the dates displayed, not the underlying 30-day calculation. A change in eligible peers can also affect the score.

Missing data stays missing

Short history, stale observations, counter resets, too few peers or a temporarily unavailable data query are shown explicitly. We do not replace these gaps with zero or draw a continuous line through them. Coverage reports the observed dates in the inclusive 31-date window, not continuous tracking. “Last observed” is a source observation date, not a claim that a collector is running now.

Limitations

Voting behaviour varies between games and portals. Public counters can change, disappear or reset; a game may receive votes without corresponding current play activity. The recorded catalogue is not guaranteed to represent every game on a portal. Changes to data coverage can alter the comparison group. Use this signal to investigate ideas, not as a substitute for direct research.

Game categories, developers and technology labels are shown only when present in collected metadata. “Similar ideas” means a shared recorded category on the same portal, not an inferred game loop or a recommendation of commercial success.

Poki/CrazyGames methodology version: votes-30d-percentile-v1.

YouTube Playables: observed play-counter growth

YouTube Playables uses a separate signal: the change in the rounded cumulative plays reported by playables.games around 30 days apart, divided by the actual elapsed UTC days. The same endpoint tolerance applies: a source observation no more than two days before each target. This is an approximate observed counter-growth rate, not an exact number of plays, unique players, concurrent users or votes. No shorter warm-up index is substituted.

The comparison group contains eligible active YouTube Playables games with valid observations, selected before checking growth. Unchanged rounded counters remain in the reference group as observed ties, but their own score is unavailable (“Rounded counter unchanged”): zero displayed change does not prove zero activity. Positive observed growth uses the same midpoint percentile formula within this separate group. Scores across portals are not comparable measures of audience size.

Missing or stale source observations, decreases, marked source corrections, removed games and insufficient history are not assigned a score. The source update date is distinct from when we fetched it: fetching old data does not make it fresh. Historical coverage begins with recorded observations; source release dates do not create historical play counts.

The upstream rounding rule is not verified. We do not infer statistical confidence or exact uncertainty bounds from a display suffix; a counter changing from 98M to 99M does not establish a million plays on that day. Source corrections and gaps can affect the result. Publisher metadata is not asserted to identify a game's developer.

YouTube methodology version: plays-30d-observed-percentile-v1. Source: playables.games catalogue.