Research Notes · Game Hype Radar

Research synthesis

What this prototype copies

The useful pattern is not a giant database. It is a ranked briefing with taste memory, clear timing, deal awareness, and fast negative feedback.

Design north star

Recommend fewer games, explain them better, and keep every signal adjustable.

Steam Discovery Queue	Small batches, ignore/wishlist/follow feedback, store preferences.
Steam Wishlist / Curators	High-intent user signal plus trusted human/editorial overlays.
OpenCritic / Metacritic	Aggregate quality, but separate score confidence from taste fit.
Backloggd / Letterboxd / Goodreads	Shelves, diaries, lists, and social proof create durable taste memory.
Deku Deals / SteamDB / IsThereAnyDeal	Historical lows, price watches, platform/store normalization.
GX Corner release calendar	Time-based browsing keeps upcoming releases visible without a huge list.
Netflix / Spotify / TikTok / YouTube	Feedback controls and novelty mixing keep discovery from going stale.
Daily briefings/newsletters	Few items, clear reason, source link, and an editorial stop rule.

Evidence ledger

How much the radar actually knows

A score is a point estimate, and not every score is equally earned. This is what stands behind them: review weight, critic coverage, how current the data is, and whether the announced date has held. Confidence never changes a rank — it sits beside it.

Median confidence 66

Evidence

Thin evidence 29

< 45

Games with history 78

304 observations recorded

Release dates that moved

银雀：十三月的爱恋 slipped 2026-06-29 → 2026-07-31 · 2026-07-20 Steam

Filtered out as pipeline events

These moved too many games on one day to be real news. A refresh changing its mind looks exactly like a wave of delays, so cohorts are labelled rather than reported.

shipped 2026-07-20 · 35
expired window cleared 2026-07-20 · 24
our data corrected 2026-07-20 · 8
shipped 2026-08-21 · 8

Ranked on little data