League History

A mobile-first web app for the ESPN fantasy league League of Extraordinary Gentlemen (league 264063) — all-time records, championships, and head-to-head.

Covers all 18 seasons, 2008–2025.

Built with Next.js as a fully static export, so it deploys free on Vercel with no server, no serverless functions, and no runtime dependency on ESPN.

Quick start

npm install
npm run dev

Open http://localhost:3000.

Deploying to Vercel

The app is a static export, so it fits comfortably in Vercel's free tier.

1. Push this folder to a GitHub repo.
2. At vercel.com/new, import the repo.
3. Vercel auto-detects Next.js. Accept the defaults and deploy.

You get a URL like league-history.vercel.app. Share it — anyone can open it, no login. To push an update, commit and push; Vercel redeploys automatically.

Add to Home Screen works out of the box via public/manifest.json. On iOS, Share → Add to Home Screen; on Android, Chrome offers an install prompt. It opens fullscreen with no browser chrome and no install warnings.

Refreshing the data

Past seasons never change, so this is a once-a-year job after a season ends:

npm run refresh

That runs two steps, which you can also run separately:

• npm run fetch — pulls the public seasons (2019+) from ESPN into data/raw/
• npm run merge:history — splits the captured 2008–2018 dump into data/raw/
• npm run build:data — transforms raw dumps into data/league.json

Then commit data/league.json and push. The raw dumps are gitignored (~12MB of mostly roster data we do not use); the derived file is 188K and is committed, because Vercel builds from it.

Merged accounts

Two managers registered more than once, and ESPN keys ownership by account GUID — so without merging, their seasons would be counted separately.

Chad Rissland has three accounts:

GUID	Handle	Seasons
{B6D2F68F…}	Youhateme39	2019–2026 (primary)
{02D15407…}	crissl1470547 / chad39	2019–2020
{0A74113B…}	ESPNfan4893280917	2021–2026

Verified they are one person rather than relatives: all three appear as co-owners of the same team (team 9) in every season they exist, never different teams. If they were different people, they would own different teams.

The merge lives in scripts/owners.mjs under OWNER_ALIASES. Because ESPN records co-ownership at the team level, stats are keyed by franchise-season and resolved to a person — summing per member GUID would have counted Chad's seasons three times. His page shows a "3 accounts merged" badge.

Steve Bazaar has two: one for 2008, another for 2009–2014. Their windows never overlap, consistent with re-registering rather than the franchise changing hands.

To merge someone else later, add their GUIDs to OWNER_ALIASES and re-run npm run build:data.

Managers who left before 2019

ESPN only returns members records for the current roster, so five managers who departed earlier have no name anywhere in the data. Their names come from the league itself and live in DEPARTED_OWNERS in scripts/owners.mjs: John Nicholson, Steve Bazaar, Matt Meyerhoff, and Ray Blakely.

What counts as what

Records are regular season. ESPN's record.overall covers only the regular season (15 weeks in recent years), while the schedule also contains weeks 16–17 playoff games. Folding those into the same totals would make head-to-head disagree with the W-L shown everywhere else, so playoff meetings are tracked separately and surfaced under each opponent on the manager pages.

The numbers reconcile exactly: 1030 regular-season game-sides + 140 playoff game-sides = 1170 = 585 games × 2.

2026 is excluded from records — the season has not been played. It is still fetched so rosters and membership stay current.

Data validation

scripts/build-stats.mjs refuses to write league.json if any invariant fails:

• league-wide wins must equal losses
• every game must resolve to a person on both sides
• head-to-head wins must equal season-record wins
• exactly one champion per completed season

These are the failure modes that would otherwise ship silently — a broken alias merge, an unresolved GUID, or head-to-head drifting out of sync.

Season coverage

All 18 seasons (2008–2025) are included, but they come from two sources with different depth:

Seasons	Source	What we get
2019–2026	Public API, fetched by npm run fetch	Full schedule, per-game scores, members
2008–2018	kona_history_standings dump, captured once by hand	Final standings only — records and ranks, no schedule

So season records, championships, playoff appearances, and finishes cover all 18 seasons, while head-to-head, per-game records, and points cover 2019–2025 only. Every page that shows a scoped stat says so inline.

Why 2008–2018 needed a manual capture

Those seasons are private (401 AUTH_LEAGUE_NOT_VISIBLE), and cookies alone do not unlock them from a script — every endpoint variant returns 401 or 404 even with a valid logged-in session. The ESPN web app reads them through:

leagueHistory/{id}?view=kona_history_standings&platformVersion=...

That exact URL returns 404 from curl with the same cookies, so the data was captured once from the browser console and committed as data/history-dump.json. npm run merge:history splits it into data/raw/ without overwriting the richer payloads for 2019+.

This is a one-time job: those seasons are finished and will never change.

Notes on the ESPN API

• Two endpoint eras: seasons from 2018 on use /seasons/{year}/segments/0/leagues/{id}; earlier seasons are documented under /leagueHistory/{id}?seasonId={year}, which returns an array of one — but see above, it does not work for this league.
• ESPN moved the host to lm-api-reads.fantasy.espn.com around April 2024. The old fantasy.espn.com host now 302-redirects API calls to a marketing page.
• 401 means the season exists but is private; 404 means no data. Do not conflate them — that is how the history first looked like it began in 2019.
• Requests need a browser-like User-Agent or ESPN rejects them.

Layout

app/
  page.js            Landing page — high-level stats
  records/           All-time leaderboards, single-game extremes
  owners/            Manager list
  owner/[slug]/      Per-manager detail, prerendered for each
  h2h/               Full head-to-head grid
components/          Nav, RankRow
lib/data.js          Data access over league.json
scripts/
  espn.mjs           ESPN API client
  fetch-history.mjs  Pull all seasons -> data/raw/
  owners.mjs         Identity resolution and alias merging
  build-stats.mjs    Raw -> league.json, with validation
  make-icons.mjs     Generates PWA icons (no image deps)
data/
  raw/               Per-season ESPN dumps (gitignored)
  league.json        Everything the app renders (committed)

Trophy Room (weekly points)

Managers earn points for weekly achievements. Both a weekly winner and a running season total are tracked.

npm run week 3      # score week 3, update data/weekly.json

Reruns are idempotent -- scoring the same week twice recomputes and replaces it, and season standings are always derived from scratch rather than incremented, so a correction can never leave a stale total behind.

Achievements live in scripts/achievements.mjs, one definition each, returning the winners for a week. Ties award everyone tied; arbitrarily breaking a tie is worse than two people sharing a badge. Points are weighted so decisions beat luck -- leading the week in raw score is +3, but leaving the fewest points on your bench is +4, because one is your roster and the other is you.

npm test exercises all 14 against a hand-built week with known answers. This matters because week 1 had no scores during development, so without it the engine's first real run would have been unverified.

Live gameday scoring

/trophies polls Sleeper from the browser every 60s and computes the same achievements client-side, so points move while games are being played.

The site is a static export with no server, which is what keeps it free and login-free. Sleeper's API is public and CORS-enabled, so the browser can read it directly -- no backend needed. Nothing is persisted from the live view; running npm run week <n> afterwards is what makes a week official.

Two consequences worth knowing:

• Streak and upset awards do not fire live. They need prior-week records, which the live view does not load. They appear once the week is scored.
• lib/achievements-client.js is generated, not hand-written. Edit scripts/achievements.mjs and run npm run sync:achievements. If the two drift, Sunday's projected points will not match Tuesday's official ones.

npm run players regenerates public/players.json, the slim name/position map the live view needs. Sleeper's full player file is ~15MB; filtering to rostered players only makes it 6KB. Re-run it after waiver moves.

Writeups

Previews and recaps are markdown files in content/writeups/, named draft-recap.md or <preview|recap>-week-<n>.md. Copy them from league-bot's out/ directory, commit, and Vercel redeploys:

cp ../league-bot/out/recap-week-3.md content/writeups/

components/Markdown.js renders them. It is a deliberately small parser covering exactly what league-bot emits -- headers, paragraphs, bold, italic, rules, lists -- rather than a dependency added to a static export for content we generate and therefore control. Extend it if the bot starts emitting tables.