How the composite score works

The idea in one paragraph

A film's composite score is built from small, checkable contributions — one for every appearance it makes on an authoritative list, in an award record, or in a major poll. A #1 placement on a heavyweight poll contributes a lot; a mid-table placement on a lighter list contributes a little; an Oscar nomination contributes a fraction of a win. But the contributions are not simply piled into one heap. They are grouped into five families of sources, and each family has a ceiling on what it can add to the total — the way a credit score caps what any single category of behavior can do to your number. A film reaches the top of this site only by earning recognition of several different kinds. Because the inputs are public facts and the arithmetic is disclosed, anyone can recompute any ranking here.

Five source families, five ceilings

We draw from five families of sources, each contributing something the others can't — and each capped, so that no single kind of recognition can dominate the score. Critics' and directors' polls (Sight & Sound 2022, in both its critics and directors editions) capture the professional consensus of the whole history of cinema at once; as the most direct measurement of all-time standing, this family carries the largest ceiling. Institutional and media canons (the AFI 100 lists, the BFI's Top 100 British films, the National Film Registry, the all-time lists of TIME, the New York Times, Empire and Rolling Stone) represent deliberate, slow canon-making, and carry the second-largest. Awards (the Oscars' Best Picture, the Palme d'Or, the Golden Lion, the Golden Bear, BAFTA Best Film, and a band of second-tier festival and academy prizes) record what juries honored in the film's own era. Community and curated canons (the Letterboxd Top 250, the Criterion Collection, directors' personal lists like Kurosawa's hundred favorites) register enthusiasm and curation outside official institutions. Annual critics' lists (Kinema Junpo's Best Ten, running since 1924; Cahiers du Cinéma's and Film Comment's year-end lists; the year-end tens of individual critics) supply a year-by-year record that catches films the retrospective polls miss, and carry the smallest ceiling.

The caps exist because signals inside a family are correlated. A film that sweeps one awards season — the festival prize, then the academy prizes, then the year-end lists — has impressed one moment in time several times over, not several eras once each. Under a plain sum, that single season could outscore fifty years of accumulated canon-making; under the caps, a sweep fills its family's quota and then stops counting. What a sweep can never do is fill the polls and canons families — only decades of retrospective judgment can do that.

Rank decay, in plain English

Within any ranked list, position matters, but not linearly — the difference between #1 and #2 is real, while the difference between #61 and #62 is noise. So each placement's contribution falls off logarithmically with rank: #1 earns full value, #3 earns about half, #15 earns about a quarter. This mirrors how canonical memory actually works: people remember what topped a list long after they've forgotten its lower half.

Winners, nominees, annual lists and memberships

An award win is an annual honor: winning the Palme d'Or in 1994 is a claim about 1994, not about all of cinema. So award wins are dampened by a constant factor before entering the score — the same factor, for the same reason, that dampens annual critics' lists, where being the best film of one year is a weaker claim than ranking high on an all-time poll. A nomination counts at a fixed fraction of a win — losing the Best Picture race is still a canonical fact, just a smaller one. Membership lists — the National Film Registry, the Criterion Collection, a director's unranked favorites — have no meaningful internal order, so every member film receives the decay value of the list's median rank: inclusion counts, position doesn't.

What the score is not

It is not a quality score, a user rating, or a prediction of your taste — and no user votes from this site are ever mixed into it. It measures one thing: how much verified canonical recognition, across how many different kinds of institutions, a film has accumulated. A beloved cult film with no institutional footprint will rank low here, and that is the correct behavior for what this instrument measures. When sources publish new editions — a new poll, a new Registry class, a new festival year — the affected placements are re-ingested and every page on the site is recomputed from the same master table.