Datawyrm

SOURCES & METHODOLOGY / 001

Keep the evidence
with the connection.

Datawyrm is a work in progress. Its first collection supports a study of game reviews and trophy data around one year after release. The atlas is an exploratory tool, not a set of completed research findings.

The research collection

The local collection contains 81 title folders, 242 CSV files, and 16,780 data rows. Of those folders, 80 contain critic, user, and trophy CSVs. Presence of a file does not establish complete coverage. Raw review text, usernames, source HTML, and PDFs are excluded from the public site.

The folder date is shown as a checkpoint. It is not assumed to be a release date or verified capture time. The import does not infer an exact anniversary, platform, edition, or observation window from the folder name. Until those fields are verified, these are exploratory sample comparisons.

What a research link means

Critic and user values are unweighted arithmetic means of valid numeric scores in the collected CSV rows. Critic scores use a 0–100 scale and user scores a 0–10 scale. Values outside those bounds or missing values are excluded from the mean; valid zero scores are retained. The row counts include every CSV data row, even a row with an invalid score. No deduplication, date-window filtering, or platform filtering has yet been applied, so these sample means must not be interpreted as official Metacritic scores.

For a selected game, critic links show up to six nearest games within five points; user links show up to six within one point. Trophy links compare CSV row counts within ten trophies, not completion rates or difficulty. Missing trophy data has no value and is not treated as zero. Ties are resolved by checkpoint date and title. Connections use only games within the current search and year filters.

The horizontal axis is the folder checkpoint date. The vertical axis changes with the selected metric. Points with missing values sit below the zero baseline. Some points can coincide; the List view provides access to every title. Color indicates which CSV types are present, not scientific data quality.

Before drawing conclusions

The SNES retail atlas

The catalog contains all 717 North American retail entries with a release in the Wikipedia SNES release table, retrieved October 6, 2026. It covers the original licensed retail era, 1991–1998. Competition cartridges, unlicensed releases, aftermarket games, and modern rereleases are outside this baseline. The source identifies North America; it does not independently establish a U.S.-only release flag for every entry.

Titles retain regional aliases, including Final Fantasy II / IV and III / VI. Publishers are selected for the North American release where regional credits are available. Day, month, and year precision are kept as supplied; an unknown day is never invented. The catalog is adapted and normalized from Wikipedia under CC BY-SA 4.0. Its cited historical Nintendo release list is a useful primary-source reconciliation target.

The Nintendo 64 retail atlas

The N64 catalog contains 296 North American retail entries from the Wikipedia Nintendo 64 release table, retrieved October 6, 2026. It covers 1996–2002, including rental exclusives where listed, and excludes 64DD, Japan/PAL-only releases and later aftermarket games such as 40 Winks. Regional publisher credits and release-date precision are preserved. This is a North American source scope, not a separate U.S.-only verification. The normalized catalog is adapted under CC BY-SA 4.0.

SNES and N64 entries keep distinct local identities. Similar names across platforms are not automatically joined as the same creative work; a future reviewed crosswalk will distinguish ports, remakes and separate adaptations.

Completion times and review scores

Direct HowLongToBeat access returned HTTP 403 during collection. The imported values therefore come from Zaire Ali’s public February 16, 2025 dataset, published February 17, 2025. These are historical observations, not current live values. The uploader lists CC BY-SA 4.0; that listing is not a verification of underlying HowLongToBeat commercial rights.

Matches require a unique normalized title or regional alias, the appropriate Super Nintendo or Nintendo 64 platform listing, and a compatible North American release year when supplied. A terminal SNES/Super Nintendo or N64/Nintendo 64 disambiguator may be stripped. Explicitly reviewed identity overrides retain their source ID and evidence note. Exact primary-title matches take priority over aliases; a failed year check cannot silently switch a title to a sequel’s franchise-name alias. Known false identity joins are excluded from displayed measurements and retained separately in the local correction audit. Fuzzy matches are not automatically accepted. Unmatched or ambiguous titles remain without imported values. Some ports have earlier cross-platform dates and are deliberately left for review.

Main-story, main-plus-extra, and completionist estimates are converted from seconds to hours, with the source sample count retained for each category. Estimates with no submissions or a zero/missing duration are represented as unavailable. These estimates combine the platforms listed on the game entry; they are not platform-specific estimates. The third-party export’s platform_* columns showed mismatched records and are excluded entirely. No game descriptions or individual player records are published.

HowLongToBeat community scores use a 0–100 scale and are included only when a positive review count is present. Nintendo Life reader ratings use a 0–10 scale. N64 reader ratings are collected under the same source and scale. A bounded selection of Nintendo Life critic scores is also included, with direct review links. Those are retrospective reviews and may discuss a Virtual Console rerelease; they are not presented as original-era reviews. Dates beside scores identify the observation/snapshot date, not necessarily the review’s publication date. Scores from different sources are never averaged into a synthetic rating.

Console connections and missing data

Publisher, developer and genre paths show up to twelve nearby releases, ordered by year. Genres are multi-label source classifications where available; series groupings remain a small editorial subset of SNES. The N64 collection does not expose a series filter until sourced groupings are available. Unknown genres or series do not create connections to other unknown entries. The map groups leading publishers and the selected game’s publisher, with remaining publishers in an “Other” lane. Within-year spacing helps separate points and does not imply exact release order.

Main-time links show up to six nearest estimates within four hours. HowLongToBeat score links use a five-point threshold; Nintendo Life reader links use half a point. The active filters constrain all links. Missing values appear below the baseline and in the List view as a dash. Coverage is shown separately for catalog entries, times, and scores: a complete title list is not complete measurement coverage.

Reading and navigating the map

The selected game’s full title stays visible. Hovering or keyboard-focusing another point reveals its full title without changing the selection. Labels are placed into available space, with thin lines linking displaced titles back to their points; lower-priority titles are hidden when they would overlap. This changes label placement only, never the underlying data coordinates.

At the overview, only the selected title and up to two connected titles are candidates for persistent labels. Zooming in admits more connected and nearby titles as space allows. Use + / − or the zoom buttons, drag to pan, or hold Ctrl/Command while scrolling. Arrow keys and directional buttons pan the view; Home or Fit all restores the overview. On touch devices, use the buttons or pinch, then drag when zoomed. Filtering or changing collection/connection type restores the overview. Selecting a game keeps the zoom and brings its point into view. The List view and details panel retain full names regardless of label visibility.

Toward a broader catalog

The next phase will add stable game identities, platform and region-specific releases, historical organization names, and sourced relationships. Acquisition should prefer permitted bulk datasets or documented APIs. Each source needs a record of its attribution, reuse conditions, rate limits, and retrieval time. Scraping, where permitted, should be cached and incremental, with a clear stop condition.

Research snapshots and catalog facts remain separate: one describes observations at a particular time; the other describes a game and its releases.

Public research summary (JSON) · SNES catalog data (JSON) · N64 catalog data (JSON) · Return to the atlas