How klieg scores work

What a klieg score is

A klieg score is the distribution of ratings from people who logged the title in their diary: the average on a 5-star scale, a histogram of every rating band, and the number of raters. We never compress it into a single percentage — one number invites manipulation; a distribution shows you the shape of opinion, including when a title splits its audience.

Scores based on fewer than 10 raters are labelled early. We’d rather show you an honest small sample than a confident-looking fake.

Who counts

Every rating is attached to a logged watch in someone’s diary — there is no rate-without-logging on klieg, so drive-by review bombing requires fabricating a watch history first. On top of that, a rating only joins the displayed score once the account behind it is at least 7 days old and has at least 10 diary entries. Watches marked private never enter a score — a small sample would otherwise reveal exactly what their owner chose to hide.

Ratings from accounts that haven’t crossed that bar yet are held, not deleted: the title page shows how many are held, and they join the score automatically the moment the account qualifies. Nothing is silently discarded, ever.

“People with taste like yours”

Signed-in members also see a second read: the same ratings, weighted by how closely each rater’s taste profile matches yours. It answers a better question than “did people like it?” — namely “did people like me like it?”. It appears once at least 3 taste-comparable members have rated the title.

Reviews and the Trusted badge

Reviews are ordered by community helpful votes, not by us. The Trusted badge is earned automatically — at least 10 helpful votes across a member’s reviews and a diary of 25+ entries — and never appointed, paid for, or revocable by favour. There is no critic caste on klieg.

Verdict labels and “% rated it 3½★+”

Alongside the distribution we show a plain-English verdict and the share of counted ratings at 3½★ or higher (the same “positive” threshold Rotten Tomatoes uses for its audience score — but we show you the derivation instead of hiding the stars behind it). Verdicts need both a share and a volume, so a label can’t be earned by three enthusiastic votes:

• Overwhelmingly loved — 95%+ positive across 50+ ratings
• Loved — 85%+ across 10+
• Well liked — 70%+ across 10+
• Divides the room — 40–70% across 10+
• Not landing — under 40% across 10+

Titles with enough recent activity also show a recent read — the same calculation over watches from the last 90 days (minimum 5) — so a show whose new season changed minds is visible as a trajectory, not hidden inside an all-time average.

Reviews written on a rewatch carry the rewatch ordinal (“2nd watch”) — conviction you can see.

The Top Rated page

Top Rated shows each title’s raw average and rater count — we never massage the number you read. The order, though, uses Bayesian shrinkage: a title’s rank blends its average with the site-wide average as if 5 extra site-average votes had been cast, so two enthusiastic ratings can’t sit above a hundred consistent ones. Titles need at least 2 counted ratings to appear, and thin samples carry the same early label as everywhere else.

What we will never do

• Sell placement, priority, or scores. No studio accounts, no sponsored surfaces.
• Quietly recalibrate a score. Any rule change lands in the changelog below first.
• Show a number we can’t explain with this page.

klieg is subscription-funded and sells nothing to studios — impartiality here is a business-model fact, not a promise. Questions? Privacy covers data handling; email support@klieg.app for anything else.

Changelog

• 2 September 2026 — Watches marked private are excluded from every score and from Top Rated. Nothing else changed.
• 21 July 2026 — Initial methodology: distribution display, early label under 10 raters, credibility hold (7 days / 10 entries), taste-weighted read (minimum 3 comparable raters), Trusted badge thresholds (10 helpful votes / 25 entries).
• 21 July 2026 — Verdict labels (volume-gated tiers above), “% rated it 3½★+” share, 90-day recent read (minimum 5 recent raters), and rewatch ordinals on reviews.
• 21 July 2026 — Top Rated page: raw averages displayed; ordering by Bayesian shrinkage (prior weight 5 site-average votes); minimum 2 counted ratings to appear. One rating per member per title — a rewatch updates your rating, it never adds a second vote.