Data Sources — oscars-award-wins
Source standards are tiered:
- Serious tier (methodology invites scrutiny): use official government or authoritative primary sources only. Crowd-edited references (Wikipedia, etc.) are NOT used — credibility is the product.
- Fun tier (low-stakes pop-culture): crowd-sourced references (fan wikis, SuperSummary, etc.) and owner-as-primary (hand-collected counts from a book or broadcast) are fine — just cite them plainly below.
This file documents every data source behind the published charts and dataset, with enough detail that someone else could independently locate and verify the original data.
Sources
DLu/oscar_data — oscars.csv
- Publisher: DLu/oscar_data (David V. Lu), compiled from the Academy of Motion Picture Arts & Sciences (AMPAS) Awards Database and IMDb.
- URL:
- Raw file: https://raw.githubusercontent.com/DLu/oscar_data/main/oscars.csv
- Repository: https://github.com/DLu/oscar_data
- Underlying primary source: https://awardsdatabase.oscars.org/
- Format: TSV (tab-separated values, despite the
.csvextension). - License: BSD 2-Clause for the compilation/code (© 2022 David V. Lu). The underlying facts are from the AMPAS Awards Database + IMDb datasets. This is a fun-tier pop-culture project; the compilation is authoritative-primary-backed, well above the fun-tier crowd-sourced bar.
- Fields used:
Ceremony(ordinal 1..98),Year(string, e.g.1927/28…2025),Class(8 broad groupings),CanonicalCategory(normalized category name across the years),Winner(boolean:Truefor a win, blank for a non-winning nomination).Category,Film,Nameare carried in the raw/export but not charted. - Coverage: 1st ceremony (1927/28) through the 98th (2025). 12,137 nomination
rows; 3,515 are wins (
Winner == True). - How the source collects the data: AMPAS publishes official nomination and winner results in its Awards Database; DLu/oscar_data parses those results and enriches each row with IMDb identifiers. One row per nomination.
- How the source defines the data: A win = one row with
Winner == True. A tie yields two winner rows (bothTrue); a multi-film / multi-nominee win is still one row.CanonicalCategorycollapses the many wording changes of the same award over the decades (e.g. the leading-acting categories) so a category reads as one continuous series;Categorykeeps the exact Oscars.org wording. - Methodology changes / series breaks: Over 98 years, categories were added,
retired, and renamed — so for any category × decade grid an empty cell means
the category did not exist that decade, NOT zero wins. Two of the eight
Classgroupings are non-competitive: SciTech (all 919 rows areWinner == True— Scientific & Technical awards have no competitive nomination, every entry is a recipient) and Special (378 of 384True— honorary / governors awards). These are excluded from the competitive heatmap (counting them as “wins” would create artificially hot cells); the six competitive classes are Title, Acting, Directing, Writing, Production, and Music. - Known controversies / debates: None material for a wins-by-category count. The only interpretive choices are the competitive-vs-honorary split and the category-normalization, both documented above and in the codebook.
- Partial current decade: the 2020s column is an in-progress decade as of the 2026-10-08 retrieval (ceremonies through the 98th / 2025 only). Its lower counts reflect fewer elapsed years, not a decline — read it as incomplete, and any published chart carrying a 2020s column should say so.
- Notes:
Yearmust be parsed to a numeric ceremony year — for a slash span like1927/28take the latter 4-digit year (1928); a plain2025stays2025.decade = (year_int // 10) * 10. Multi-value fields (Nominees, NomineeIds) are pipe (|)-separated in the raw file. - Retrieved: 2026-10-08
The Movie Database (TMDB) — film genre enrichment
- Publisher: The Movie Database (TMDB), https://www.themoviedb.org/
- Required attribution: “This product uses the TMDB API but is not endorsed or certified by TMDB.” (TMDB’s attribution requirement; carried in every chart subtitle that uses genre and in the export codebook.)
- URL: API base https://api.themoviedb.org/3 (endpoints
/3/find/{imdb_id},/3/search/movie,/3/genre/movie/list). - Format: JSON over the TMDB v3 REST API. An API key is required and is kept out of version control (never printed, logged, or committed). Per-film responses were cached locally so re-runs are offline.
- License: TMDB API Terms of Use — free for non-commercial use with the attribution above. The genre labels are TMDB’s.
- Fields used:
primary_genreonly — the first genre of the matched movie (genres[0], i.e. the first entry of TMDB’sgenre_ids). The full genre list andtmdb_idare kept in the cached JSON for provenance but do not flow downstream. - Coverage: looked up for the unique films that won a MAJOR award (see the major-award definition below) — 464 distinct films — not all 3,515 wins. A film that won several majors is fetched once; its genre is joined back onto each of its winning rows.
- How the source collects the data: TMDB is a community-maintained movie database; genres are assigned by TMDB’s contributors/editors from a fixed TMDB genre taxonomy (Drama, Comedy, Action, Romance, etc.).
- How the source defines the data: “Primary genre” here is TMDB’s first
listed genre for the film (
genres[0]). TMDB lists multiple genres per film; we deliberately keep only the first (owner decision: primary genre only). This is a single, reproducible label, not a judgment about the film’s “true” genre. - Match method: the IMDb id carried by the Oscars dataset (
FilmId, alreadytt-prefixed) is matched to TMDB with/3/find/{FilmId}?external_source=imdb_idas the PRIMARY, deterministic match (no title ambiguity). A title + ceremony-year search (/3/search/movie) is used only as a FALLBACK when/findreturns no movie result. For the 2 major rows whoseFilmIdis pipe-separated (multi-film award), the first id is used for the genre lookup (the award is one win; genre is attributed to the primary/first film). - Methodology changes / series breaks: TMDB genres are present-day labels from the current TMDB taxonomy — they are the modern genre classification, not the contemporaneous (release-era) marketing genre. A 1930s Best Picture carries the genre TMDB assigns it today. Treat the genre as a consistent modern lens across all decades, not as how the film was marketed at the time.
- Known controversies / debates: genre is inherently fuzzy and a film often spans several; using only the first TMDB genre is a simplification. Many winners are legitimately multi-genre — e.g. The Silence of the Lambs is commonly tagged Crime/Thriller/Horror and Oppenheimer Biography/Drama/History — but each is charted under its single TMDB primary genre only. This caveat is carried in the README prose above the charts so a reader isn’t surprised that a film they think of as another genre shows as Drama. Rare genres are binned into “Other” for legibility (see codebook).
- Notes:
primary_genreis populated for major-award films only; it is NULL for non-major wins in the export (documented in the codebook). - Retrieved: 2026-10-08
Major-award definition (used for the genre enrichment)
“Major awards” = the eight headline CanonicalCategory values, enumerated from
the data (756 winning rows total):
| CanonicalCategory | Class | Wins |
|---|---|---|
| BEST PICTURE | Title | 98 |
| DIRECTING | Directing | 97 |
| ACTOR IN A LEADING ROLE | Acting | 99 |
| ACTRESS IN A LEADING ROLE | Acting | 99 |
| ACTOR IN A SUPPORTING ROLE | Acting | 90 |
| ACTRESS IN A SUPPORTING ROLE | Acting | 90 |
| WRITING (Adapted Screenplay) | Writing | 98 |
| WRITING (Original Screenplay) | Writing | 85 |
Deliberately excluded (not “major” in this sense): DIRECTING (Comedy/Dramatic
Picture) (1920s one-offs superseded by DIRECTING), ASSISTANT DIRECTOR,
WRITING (Original Story) and WRITING (Title Writing) (early/auxiliary writing
awards — Original/Adapted Screenplay are the modern screenplay majors), and every
craft (Production), Music, short-film, documentary, and honorary category. This
boundary is applied consistently across the whole dataset.
Notes on Data Quality
- Original source files are preserved verbatim and never modified in place.
- Series breaks: over 98 years the Academy added, retired, and renamed categories, so an empty category × decade cell means the category did not exist that decade, not zero wins. The 2020s is an in-progress decade (data through the 2025 ceremony).
- Definitions drive comparisons: the win count, the competitive-vs-honorary split, the category normalization, and the single-first-TMDB-genre choice are all spelled out above and in the codebook, so the numbers in the charts are defined the same way across every decade.