MetaLens VGC

An interactive, cross-filtering metagame dashboard for Pokémon Champions, the VGC 2026 format (regulations M-A, M-B and M-C). It's a free static site built for Cloudflare Pages at https://metalensvgc.pages.dev.

Every number on the site is aggregated in your browser from real tournament team sheets and ladder statistics fetched by the data pipeline in scripts/. Nothing is invented or "calibrated". When a filter leaves too little data, the widget says Insufficient data instead of guessing.

Features

• Global filter bar (sticky): regulation (M-A / M-B / M-C, defaults to the current one), source (Tournaments / Smogon Ladder / Ranked in-game), event tier (Worlds / Internationals / Regionals / Online), placement (All / Top Cut / Top 8 / Winners), date range and a minimum-sample slider.
• Cross-filtering everywhere. Click a Pokémon, type, archetype, item, move, Mega, core or team to add a filter chip, and every chart recomputes. Type, item and move chips describe a Pokémon: with Electric selected, the leaderboard, types, items, speed tiers and trends count only Electric Pokémon (and "Electric + Life Orb" means an Electric Pokémon holding Life Orb), while teammates, archetypes and the team library show those whole teams. Pokémon, core, Mega, archetype and team chips pick teams; an archetype chip means teams whose main archetype it is (the donut slice you clicked). Shift-click or Alt-click excludes instead ("teams without X"). You can remove chips one by one or clear them all.
• Shareable state. All filters, chips, the skin and the theme live in the URL hash.
• Hover links. Hovering a Pokémon highlights it across charts.
• Sections
  • Meta Snapshot
  • Usage leaderboard with 95% Wilson intervals
  • Type landscape: type usage, and which move types the field is weak to, by real multiplier (4×, 2×, ½×, ¼×, 0×)
  • Item usage: which held items the field runs (share of teams, win rate, who holds them; Mega Stones optional) and the most common items for every Pokémon, searchable. The leaderboard also shows each Pokémon's top item under its name. Hand-typed sheet spellings ("focus sash", "U-Turn", "None") are normalized to dex names when the data loads.
  • Archetypes: classification, split, and a matchup heatmap built from real pairings
  • Usage vs. win-rate quadrant
  • Teammate co-usage (raw % / lift), top cores, and teammate rate for any Pokémon
  • Speed tier matrix with Tailwind, Scarf, ±1, paralysis and Trick Room toggles, plus a "my Pokémon" benchmark
  • Weekly trends with a brush that drives the date filter, risers and fallers, and the regulation shift
  • Team Sheet Library with one-click Showdown paste copy
  • My Team Scanner: paste a team to see its weaknesses, unanswered threats and speed position, plus evidence from real matches of similar teams: best and worst matchups (Pokémon and archetypes), an item check (with item clause), common teammate picks and the weakest link
  • Pokémon deep-dive drawer
  • Credits (made by Vin: vwesley.dev, Instagram @vinnql, X @Vin_Koe)
  • Data & methodology, with JSON/CSV download
• Two skins, green palettes: Pro (default: JetBrains Mono and Inter, the palette of vwesley.dev) and Retro (Press Start 2P, Silkscreen, VT323, CRT scanlines, pixel borders, Game Boy greens). Each comes in light and dark; dark is the default. A faint animated pixel field sits behind the page (still under reduced motion).
• Every chart says what a click does in a one-line hint under it. The layout is responsive down to 390px.
• Active filters always visible. Chips sit in a highlighted bar pinned under the filter bar, with a count and a Clear all button, so you always know what the page is filtered by.
• Collapsible filter bar. "▲ Hide" (or the F key) folds the filters and the header settings into a one-line summary; the choice is remembered per browser. Small screens start collapsed.

Champions mechanics used

• No Tera. Mega Evolution is the gimmick, so the Mega choice appears where a Tera widget would normally go.
• Stat Points (SP) replace EVs/IVs. Each Pokémon gets 66 SP, with at most 32 in one stat. At Level 50:
  • HP = base + 75 + SP
  • every other stat = floor((base + 20 + SP) × nature), where nature is 1.1, 1.0 or 0.9
  • This is implemented and unit-tested in public/js/lib/stats.js. For example, Jolly Garchomp with 32 Speed SP has 169 Speed.
• Usage is keyed by display form. A Pokémon holding its own Mega Stone counts as the Mega form (Charizard-Mega-Y, Floette-Mega, …), the same way Smogon and Pikalytics count it. Gendered and regional forms stay distinct (Basculegion / Basculegion-F, Indeedee / Indeedee-F, Floette-Eternal / Floette-Mega).

Data sources

Source	Used for	Notes
Limitless Play API (/api/tournaments, /standings, /pairings)	Online tournament teams (open team lists), W-L records, top cut, match results	Formats M-A, M-B, M-C. Public API with a 50 requests / 5 min limit
Limitless VGC + standings.limitlessvgc.com	Official events (Regionals, Special Events, Internationals, Worlds): published team lists, records, top cut, pairings	RK9 team lists are disallowed by RK9's robots.txt, so RK9 is not used
Smogon usage stats (chaos JSON, gen9championsvgc2026reg{ma,mb,mc}-1760)	"Ladder" source: usage, items, abilities, moves, SP spreads, teammates	Kept separate from tournament data. Cutoff 1760, non-Bo3 ladder
Pokémon Champions Battle Data (/api, /api/battle/Doubles/<id>?season=)	"Ranked (in-game)" source: the official in-game ranked ladder's Battle Data (usage rank, and per-Pokémon moves, items, abilities, natures, SP spreads, teammate order)	Doubles only, last snapshot of each in-game season (M1–M2 = M-A, M3–M5 = M-B, M6 = M-C). The game publishes ranks, not usage shares, so none are shown. Battle data provided by Pokémon Champions Battle Data, used under its API rules (attribution; no redistribution as a data service)
pokemon-showdown npm package (champions mod)	Dex: types, base stats, abilities, move data, Mega Stones	Covers every Champions and Z-A Mega, so no override table was needed
Pokémon Showdown sprites

Pokémon and its characters and sprites are trademarks and copyrighted works of Nintendo, Game Freak, Creatures Inc. and The Pokémon Company. Sprites are hot-linked from Pokémon Showdown's sprite server (the Showdown project does not own them). MetaLens VGC is an unofficial, non-commercial fan project and is not affiliated with them. The data belongs to its sources. This repository's code is MIT-licensed (see LICENSE); third-party components and data are listed in THIRD_PARTY_NOTICES.md.

Data and privacy

• Player names and countries come from public tournament results (Limitless, official event pages) and are credited above. To request removal of your name, open an issue or contact the owner through GitHub (@vincentwesley).
• No analytics, no cookies, no accounts.
• The site only calls its own origin and the sprite host (plus Google Fonts for typefaces).

Decisions and known gaps

Decisions made while building, and the gaps they leave:

• Which online events count. A Limitless Play tournament is included when it has ≥16 players, open decklists, and its date falls inside the regulation's window. The regulation is decided by date, not by the organizer's format label, so events tagged with the wrong format are dropped. Events less than 48 hours old are skipped as possibly still running. Events from the last 7 days are refetched on every run.
• Official events come from limitlessvgc.com: Indianapolis, Turin and NAIC (M-A), Worlds (M-B), and Baltimore, Frankfurt and Brisbane (M-C). Records, top cut and pairings come from the standings pages on standings.limitlessvgc.com. RK9 is not used because its robots.txt disallows the team-list pages.
• Tiers. Special Events and Japanese events count as Regionals. NAIC counts as an International.
• Top cut. Online top cut is the set of players in bracket (phase ≥ 2) pairings. Official top cut comes from the standings' top-cut flags. When the cut is unknown, top 8 counts as top cut.
• Win rates come from each team's standings record (W-L-T). Matchup cells come from real pairings where both players have a team sheet in the data.
• No Stat Points from tournaments. Limitless doesn't publish SP spreads, so no tournament team has one. Spread and speed data comes from the Smogon 1760 ladder where it exists. Otherwise the site shows theoretical min–max speeds, labelled as such.
• Ladder gaps. Smogon's "Checks and Counters" tables are empty for these formats, so that widget shows an explicit empty state. M-C has no ladder month published yet, so the Ladder source is unavailable for M-C.
• Mega keying. A Pokémon holding its own Mega Stone counts as the Mega form. A few sheets list two Mega Stones; every one of them is counted (team.megas).
• Indeedee. A few official-event team lists give plain "Indeedee" with no gender. These are counted as Indeedee-F, since the female form is the one played in doubles. Online Limitless sheets are taken as listed.
• Archetypes are rule-based: each archetype's rules live in public/js/lib/archetypes.js. A team can match several archetypes; the first match is its primary one.
• "vs previous period" deltas compare against a window of the same length immediately before the selected one, inside the same regulation. If that window starts before the regulation does, the comparison is against the whole previous regulation. The previous sample size is shown, and deltas from small samples are flagged.
• Ranked (in-game) source. This data covers individual Pokémon, not teams, so the team-level views (types, archetypes, quadrant, trends, library, scanner) say they are unavailable instead of guessing. Speed tiers prefer a real source in this order: team-sheet SP, then the ranked ladder's most common spread with its most common nature, then the Smogon ladder, then theoretical bounds. The game reports spreads and natures separately, so pairing the top spread with the top nature is an approximation, and the chart labels it as one.

Data coverage

Numbers from the build on 2026-09-29. Online teams are events with ≥16 players; official teams are published top lists.

Regulation	Window	Events	Teams	Official teams	Matches	Ladder (Smogon)	Ranked (in-game)
M-A	Apr 8 – Jun 17	328	24,057	436 (Indianapolis, Turin, NAIC)	64,335	Apr–Jun	seasons M1–M2
M-B	Jun 17 – Sep 9	464	33,316	57 (Worlds)	85,114	Jun–Aug	seasons M3–M5
M-C	Sep 9 – now	59	3,576	365 (Baltimore, Frankfurt, Brisbane)	8,953	none yet (September stats land in early October)	season M6 (daily)

No tournament team has an SP spread, because Limitless doesn't publish them. Natures are known for 16% of M-A mons (early Limitless sheets omitted them) and for 100% in M-B and M-C.

How the tournament numbers compare with the reference usage figures from the brief (Pikalytics-style):

• M-C. The 8 reference Pokémon are the same 8 at the top of the in-game ranked ladder, in nearly the same order. Our tournament sample agrees on Sneasler (38.7% vs 36.6%), Incineroar (30.9% vs 28.5%), Salamence-Mega (27.7% vs 26%) and Kingambit (22.2% vs 25%). It runs much higher on Rillaboom (54.7% vs 37.6%) and lower on Basculegion (13.8% vs 23.8%), Golisopod (13.0% vs 21.7%) and Indeedee-F (17.4% vs 21.6%). The M-C sample covers only about two and a half weeks of mostly-online events, where the Rillaboom-plus-double-Fake-Out core is even more dominant than on the ladder. The site's Ranked (in-game) source shows the ladder view next to it.
• M-A. Garchomp (39.3% vs 40.5%), Sinistcha (25.1% vs 20.4%) and Whimsicott (15.2% vs 17.7%) are close. Basculegion (34.1% vs 51.5%), Kingambit (33.8% vs 40.7%) and Floette-Mega (14.1% vs 23.9%) are lower. 57% of our M-A teams come from April, the launch month, when Basculegion was at 27%. It climbed to 45.5% in May, reaches 43.6% among top-cut teams and 55.5% among official-event teams. The reference numbers track that later, more competitive meta. The Top Cut filter and the date brush on the site reproduce these slices.
• M-B. Charizard-Mega-Y and Garchomp are both in the top 6 (25.3% and 32.1%), consistent with the "Big Six" reference.

Local development

npm ci
npm run data     # full pipeline (incremental; first run takes ~3 h due to Limitless rate limits)
npm test         # node:test unit tests
npm run serve    # http://localhost:8080
npm run e2e      # Playwright checks against the local server
npm run e2e:cloud # same checks with a preinstalled Chromium, third-party hosts stubbed (offline sandboxes)
npm run shot     # screenshots of both skins, desktop + mobile, into screenshots/

Useful pipeline flags: node scripts/build-data.js --reg M-C (one regulation) and --max-tournaments N.

How the data refresh works

.github/workflows/refresh-data.yml runs every Monday at 00:00 UTC (08:00 GMT+8), and you can also start it by hand: Actions → Refresh data → Run workflow.

What it does:

1. Checks out the branch, restores the data-raw/ HTTP cache, runs npm ci, npm run data and npm test.
2. Commits and pushes public/data/ (if anything changed), which triggers a Cloudflare Pages redeploy. Before pushing, the data commit rebases onto the branch's latest tip (with retries). Other commits pushed during the run don't cause failures; runs never overlap.

The pipeline is incremental: it reuses every completed tournament and finished ranked season already in the committed public/data, refetches online events from the last 7 days (they may have been captured mid-event), skips events under 48 hours old, and rate-limits requests per host with a descriptive User-Agent.

Failure handling:

• If any step fails, a github-script step opens an issue titled "Weekly data refresh failed" (or comments on the open one) with the run link, trigger name, failing step, and the last 40 log lines of that step. Nothing is committed on failure, so the site keeps serving the last good data.
• A hang fails the build step at 300 min, allowing the alert to still run (a job-level timeout would cancel it).
• On the next successful run, the issue is commented "Refresh succeeded in " and closed.
• The repo owner is notified of new issues by default. To ensure notifications reach you, go to GitHub → Settings → Notifications, enable "Issues" for email and/or GitHub Mobile push, and "Watch" the repo (at minimum "Issues"). Scheduled workflows are disabled after 60 days without activity; the weekly data commit counts as activity, but if the site goes quiet, re-enable it in the Actions tab.

Data staleness:

• When manifest.generated is more than 10 days old, a notice under the page header says "Data last refreshed — the weekly update may have failed; figures may be out of date." (lib/stale.js; an e2e check verifies this notice).

Pipeline robustness:

• A new in-game ranked season that starts after the last known regulation (scripts/lib/regs.js) is skipped with a warning instead of failing the run. When a new regulation is announced, add it to scripts/lib/regs.js (and update SUFFIX in scripts/sources/ladder.js, and manifest.current in scripts/build-data.js).
• Smogon's monthly stats (e.g., 2026-09) land in early October; the pipeline automatically tries each month up to the current one and skips 404s. Schema validation failures still fail the run.

Performance:

• From an empty data-raw/ cache on GitHub's runner: ~47 min total (official events ~21 min, Limitless online ~15 min, ranked M-C ~5 min, sprite checks ~6 min). No host blocked or rate-limited the runner.
• The cache (~5 MB) is saved only after a successful run.

See docs/TESTING.md for test coverage and what has been verified.

Deploying on Cloudflare Pages

1. Cloudflare dashboard → Workers & Pages → Create → Pages → Connect to Git.
2. Pick the GitHub repo vincentwesley/MetaLensVGC.
3. Project name: metalensvgc, which gives the site https://metalensvgc.pages.dev.
4. Production branch: the branch you deploy from (merge this branch into main, or pick this branch).
5. Framework preset: None. Build command: leave the field blank (don't type anything; if it can't be saved blank, use exit 0). Build output directory: public. Optional: environment variable SKIP_DEPENDENCY_INSTALL = 1 skips the unneeded npm ci.
6. Save and Deploy. Every push to the production branch redeploys, including the weekly data commits.

public/_headers sets the cache and security headers (a CSP that allows only this site, Google Fonts and Pokémon Showdown sprites).

Project layout

See CLAUDE.md for conventions and module contracts, and public/data/SCHEMA.md for the data contract.