Hype Climb

How the hypemeter works

Every tracked game gets a hypemeter score from 0 to 100 -- our best guess, right now, at how anticipated it is. It's not a review score and it isn't trying to predict sales; it's a single number built from 11 signals we can actually measure, blended together with fixed weights, plus a bonus for a game continuing a well-loved series.

The signals

On the game page, both meters draw a cool segment beside the warm fill: its width is exactly what recent pulls have cost the score tonight. The chips beside the events row name each recent event and its own direction -- a plus for a push, a minus for a pull -- and a delay or an earlier move graded slight, moderate or major carries a label naming that grade, because a major one counts for more than a minor one.

Each signal is saturating: past its threshold, more of the same input keeps helping but with steadily diminishing returns, so one outlier number (a game with thousands of IGDB hypes, say) can't blow the scale out for every other game. A game with no data yet for an external source (no Wikipedia article, no known videos, no YouTube key configured) simply contributes 0 for that row -- the row still shows, so the score never hides which signals it does and doesn't have.

The 11 weighted signals are combined as a weighted average and scaled to 0-100: each signal's saturated value (0 to 1) times its weight, divided by the total of all the weights (19). Series pedigree sits outside that average, and is the one part of the score that is purely additive: it adds up to 15 points on top, and adds exactly nothing to a game without one. That matters more than it sounds -- it means crediting sequels can never quietly push every other game down the rankings to make room. You can see this same math broken down per-game on any game page, under "Why this score?".

A game page also draws the recent history of that score as a small line. Read it as a record of the number we computed, not as a measure of anticipation rising or falling: the formula changes as signals are added or re-weighted, and when it does many games' scores move that night — so a step in the line can be a change in how we score rather than a change in how anticipated a game is.

That line also carries a small dot on any day something happened to the game — a delay, a new trailer, a platform appearing. A dot states when something happened, nothing more: events are one of 11 signals feeding the score, and one of the smallest, so a dot sitting next to a step in the line does not mean the two are connected.

Most of what moves this score is gradual and carries no event at all, so a stretch of line with no dots is not a quiet stretch — it can be exactly when the score is changing fastest.

Each night's stored score records which formula version produced it; the version in force today dates from 2026-09-01b. A game page's weekly-change line compares only two readings of the same version, over the trailing 7 days -- a step across versions renders nothing at all, rather than a false story about anticipation moving.

Hype freezes at release

A game's hype score is a pre-release quantity. It stops moving the day a game releases or is marked cancelled: the score freezes at whatever it reached, we stop collecting the external signals above for that game, and the frozen number stands as a permanent record of how anticipated it was -- not a live figure that keeps decaying afterwards.

A second score for a late port

A game's hype freezes at release, but a dated port that lands on another platform much later earns its own second score: the same signals above, with the proximity component re-anchored to the port's date and the YouTube and showcase-exposure components re-scoped to count only activity from the port era onward, not the original release's. It updates nightly while the port is still upcoming, and freezes the moment the port arrives.

How much of the score actually comes from IGDB

The hypemeter is Hype Climb's own score: we compute it ourselves from the signals above, not from anything IGDB calculates. IGDB's own catalogue still runs through more of the meter than any other single source -- IGDB anticipation, anticipation growth, showcase exposure and want-to-play momentum all draw on it directly, and release proximity draws on it indirectly, through the release dates IGDB supplies -- but no single IGDB signal leads the meter any more: those four IGDB-sourced signals together carry 8 of the 19 total weight, and the largest of them now sits level with video cadence -- our most independent signal -- rather than ahead of it. Wikipedia pageviews and edits, and YouTube's view counts and upload cadence, are our ongoing step away from leaning on one enthusiast catalogue: Wikipedia adds a mainstream-attention check, and YouTube adds a video-reception check -- though YouTube stays optional (it needs an API key we haven't wired up in every environment) so its absence never breaks the score. The weights above are tunable constants, not settled science, and we expect to keep shifting them as Hype Climb's own readership grows and the newer signals accrue more history of their own.

Demand and Campaign

Beside the meter, every scored game page shows two readings built from the same signal rows: Demand — what an audience does (IGDB anticipation, trailer views, Wikipedia attention and edits) — and Campaign — what a publisher does (video cadence, showcase appearances). They are separate readings of the signals, not the score split in two: several signals sit on neither axis, the two are on different scales, and neither adds up to the hype score. We publish them because one number cannot say what they can — a game can be intensely wanted while its publisher is quiet, and the most anticipated game in the catalogue has spent months exactly there. More broadly, that is what the hypemeter measures: behaviour around a game — what its audience does and what its publisher does — never the size of any one site's crowd.

The four kinds of attention

Each phase falls into one of four cells, measured against the tracked catalogue's own median on each axis, never a fixed number, wherever both readings clear the floor: Broad — above the typical tracked game on both; Audience-led — above it on demand, below it on campaign; Publisher-led — above it on campaign, below it on demand; Low profile — below the typical tracked game on both. An unscored phase, or one whose demand or campaign reading is too thin to compare, gets none of the four — and neither does a frozen phase, since the comparison is against games still being measured. The Attention page ranks the games whose audience is furthest ahead of their publisher's campaign.

What Low, Moderate, High and Extreme mean

The hypemeter score itself is banded into four levels, re-fitted to the real spread of scores across the catalogue rather than a fixed quarter each: a score below 20 reads Low, 20 up to 32 reads Moderate, 32 up to 58 reads High, and 58 and above reads Extreme — deliberately the rarest word on the site.