Methodology
A daily read on attention—not a verdict.
Hyped24 combines several kinds of recent, attributable public signals into one experimental index. It is designed to make changes in attention easier to notice while staying honest about missing or aging information.
Four views of attention
Conversation and audience carry the greatest emphasis. Momentum and declared interest add context when trustworthy evidence is available.
Primary signal
Conversation
How much fresh, attributable discussion and publishing activity surrounds a game.
Primary signal
Audience
The size of the audience currently watching, viewing, playing, or otherwise paying attention.
Supporting signal
Momentum
Whether comparable conversation and audience signals are accelerating or cooling over time.
Supporting signal
Declared interest
Verified public evidence of intent, such as following, wishlisting, or preordering when available.
Comparable, not absolute
Signals are considered in the context of the tracked catalog so one platform or naturally large number does not dominate every game.
Missing is not zero
A confirmed zero is real information. An unavailable or failed source is not silently turned into zero; it reduces coverage and confidence instead.
Smoothed, still responsive
Scores are produced for completed UTC days and gently stabilized against recent history. They can still move sharply when sparse or fast-changing provider inputs move.
Mood is separate
A loud conversation can still feel mixed.
Mood summarizes the direction of available sentiment-like signals. It never raises or lowers the attention score, and it remains unavailable when there is not enough trustworthy evidence.
How to read a score
Confidence reflects how much of the expected signal picture was actually available. It is possible for a high score to have modest confidence, so the two should always be read together.
Freshness shows when the score was computed and which source periods it represents. The last successful score stays visible through collection failures and is clearly marked stale after 48 hours.
High confidence
At least 80% of the expected signal picture was available.
Medium confidence
Between 55% and 79% of expected signal coverage was available.
Low confidence
Between 30% and 54% was available. Below 30%, or with fewer than two external providers, no score is published.
Calibrated experimental version
attention-v2 is frozen for the public beta.
The launch decision was recorded after reviewing seven closed UTC dates from 8 August 2026–14 August 2026. This version uses corrected collection windows and completeness rules. Earlier attention-v1 snapshots remain immutable archived evidence and are never used to smooth, rank, or calculate v2 movement.
Frozen means its weights, thresholds, normalization, and input rules will not be edited in place. A later change requires a new named version and cannot rewrite existing history.
Limits to keep in view
Relative and volatile
Scores are relative to the currently tracked catalog. Daily YouTube samples, low-volume Twitch categories, momentum, and catalog percentiles can produce substantial short-window movement even when every input is valid.
Some games remain unscored
Missing mappings, unavailable providers, incomplete collection, and capped result sets are not guessed or converted to zero. If fewer than two compliant external providers remain, Hyped24 publishes no score.
The 30-day history is still building
attention-v2 begins on 8 August 2026. Until 30 closed v2 dates exist, the 30-day chart is necessarily incomplete and retains honest calendar gaps rather than borrowing or backfilling v1 scores.
Negative-event validation continues
Calibration confirmed that mood is calculated independently from attention, but the seven-day window contained no clean live negative-attention case with sufficient provider coverage. The first suitable post-launch case will be reviewed before the experimental label is reconsidered.