아직 감지된 업데이트가 없습니다. GitHub 릴리스가 나오면 여기에 보입니다.
최근 완료된 8시간 회차 사이 GitHub 스타 증가량 순
Use Zwift Click, Zwift Play, Shimano Di2, or any Bluetooth controller with MyWhoosh, Rouvy, Zwift, and more. Free to try on iOS, Android, Windows, and macOS.
개인 계정의 공개 저장소 소유자는 GitHub로 로그인해 운영자를 확인하고 소개를 직접 관리할 수 있습니다. 관리자 승인은 필요하지 않습니다.
로그인한 뒤 확인 파일을 만들어 이 저장소의 기본 브랜치에 30분 안에 커밋해 주세요. 다른 브랜치나 포크의 파일은 인정하지 않습니다. 파일을 올린 뒤 아래에서 확인하면 관리자 승인 없이 관리할 수 있습니다. 자격을 유지하려면 파일을 남겨 두세요.
로그인이 필요하면 로그인 화면으로 이동합니다. 로그인 후 이곳으로 돌아와 원하는 확인 버튼을 다시 눌러 주세요.
프로젝트 폴더에서 /nomorevibe verify를 실행하면 소유권을 확인하고 소개를 직접 관리할 수 있습니다.
1분 만에 확인하기README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
The current application is hosted on Vercel at https://going-long.vercel.app/.
It includes football betting/fantasy, baseball integration, server-side odds
routes, a private React validation journal, and a separate local Python scanner.
Independent reviewers, including Claude: start with CLAUDE.md
and the audit guide for the current code map and tests.
The original setup notes below predate the current hosting, API and database
architecture. Their GitHub Pages/no-server instructions are historical, not the
current deployment procedure. See vercel.json and the audit guide instead.README 전문 보기 ↗The current application is hosted on Vercel at https://going-long.vercel.app/. It includes football betting/fantasy, baseball integration, server-side odds routes, a private React validation journal, and a separate local Python scanner.
Independent reviewers, including Claude: start with CLAUDE.md and the audit guide for the current code map and tests.
The original setup notes below predate the current hosting, API and database
architecture. Their GitHub Pages/no-server instructions are historical, not the
current deployment procedure. See vercel.json and the audit guide instead.
A draft war room built on the Late-Round Draft Guide's model. Every player is priced in expected PPR points per game, valued against your league's actual replacement level, and scored by what it costs you to wait.
Runs on GitHub Pages. No build step, no server, no API keys.
You haven't pushed anything yet, so start here.
1. Create the repo
cd path/to/going-long
git init
git add .
git commit -m "Going Long: draft board"
git branch -M main
git remote add origin https://github.com/YOUR_USERNAME/going-long.git
git push -u origin main
2. Let Actions write to the repo
Settings → Actions → General → Workflow permissions → Read and write permissions → Save. The data job commits back to the repo, so it fails without this.
3. Turn on Pages
Settings → Pages → Source: Deploy from a branch → branch main, folder
/ (root) → Save. Your board lands at
https://YOUR_USERNAME.github.io/going-long/.
4. Build the data
Actions → Refresh draft data → Run workflow. Takes about two minutes and
commits data/players.json plus nine format files. After that it runs every
morning on its own.
If none of your leagues are 12-team, edit the FORMATS list at the top of
scripts/build_data.py before running.
5. Connect your leagues
Open the page. Enter your Sleeper username → Find my leagues → pick the league and draft → Add this draft. Repeat for every league you're in; the dropdown in the header switches between them. Team count, scoring, roster slots and superflex are all read off Sleeper — nothing to configure by hand.
6. Load the guide
Under Rankings & tiers, select all five LateRoundQB CSVs at once
(ppr, half-ppr, te-premium, superflex, superflex-te-premium). Each
one is stored as its own board and tagged by format from its filename, and
every league you add picks up the board matching its shape automatically — a
superflex TE-premium league gets the superflex TE-premium rankings without you
choosing. Pasted cheat sheet rows work too.
Under Market Score, paste all four position tables at once. Anything that doesn't name-match gets a dropdown to fix — usually five or six players.
Both live in browser storage, so they persist without being committed. If you'd
rather keep them in the repo, export from devtools and drop the JSON in
data/ — nothing in the app cares which.
7. Optional: profile your league
LEAGUE_ID=your_league_id python scripts/manager_profiles.py
Walks every draft your league has ever run and writes data/managers.json —
who reaches, who waits, who drafts straight off a sheet. Not wired into the
board yet.
EDGE is what the current pick buys you over waiting. Not "how good is he" — if a player you love won't be taken for another twenty picks, waiting is free and his edge is low. It spikes when someone unlikely to survive is also well clear of the next man at his position.
VORP is points per game above your league's replacement level, which is derived from your actual starting slots. In a superflex league the QB baseline slides from QB12 to roughly QB20 and every quarterback's VORP jumps accordingly — nothing about that is hardcoded.
MS is Market Score, converted into points. The model's ranking is compared to the ADP ranking, and the gap between what those two draft slots have historically been worth is the edge.
The survival bar under the edge number is the chance he lasts to your next pick. Teal is safe to wait on, red means gone.
Tags on the name line:
NEED — the quality supply at that position won't cover what you still have
to fill by your pick after next.
RUSH 8.9 — a top-8 quarterback who ran for that many fantasy points per game
last year. Four or more is the bucket that hits.
POCKET — a top-8 quarterback under two rushing points per game. Historically
the worst early-round bet at the position.
N-1 9.3 — a top-6 tight end who missed 12 PPR points per game last season.
TIER THIN — fewer players left in that tier than picks before your turn. It
will not survive.
COMES BACK — the tier is deep enough that someone from it returns to you.
Take the scarce tier instead.
FIRST QB / FIRST TE — nobody has taken one yet. Being first off the board
at a onesie position risks paying a premium the room was never going to pay.
The sliders are Rank (how much the guide's board overrides raw ADP), MS (how much Market Score divergence counts), and Bench (what a player who doesn't crack your lineup is still worth). Defaults are .60 / .40 / .30.
The Tier watch panel shows every live tier's remaining count against your pick gap, which is the guide's tiering rule as a running readout.
| What | Where | Notes |
|---|---|---|
| Players, drafts, picks, rosters | api.sleeper.app/v1 | No auth, CORS-friendly, read-only. |
| Consensus ADP + stdev + byes | fantasyfootballcalculator.com/api/v1/adp | Pulled server-side in the Action. |
| Market values | api.fantasycalc.com/values/current | Carries sleeperId, so the join is exact. |
| Expected points, replacement level, tiers, Market Score | The guide | Curves and tables are in docs/MODEL.md; rankings you import. |
| Prior-season production | nflverse + DynastyProcess IDs | Weekly stats aggregated to per-game rates, joined on Sleeper ID. |
The board polls Sleeper every 3 seconds while the draft is live and stops when the tab is hidden. Everything else is a static file.
my_gain × their_perceived_gain.docs/BETTING_MODEL.md for the implemented statistical contract and remaining validation limits.The board opens directly with NFL props and a separate NFL/NCAA game-lines view. Player projections use trailing nflreadpy game statistics, a nonpositive-mass lognormal yardage model, and Poisson receptions/TD counts. The board displays model win probabilities, push-aware EV and capped quarter-Kelly allocations. It preserves bookmaker/line identities and flags missing or stale data.
The nightly workflow now runs scripts/build_pipeline.py after odds ingestion,
writing configuration and projections into data/data.json and
data/history.json. Set the Actions repository secret PARLAY_API_KEY to
enable prop quotes. Historical model rebuilding does not need an odds key.
The hosted board refreshes NFL props and NFL/NCAA game lines through the
server route /api/odds. Set the same PARLAY_API_KEY secret in hosting.
Run npm ci and npm run dev locally with the key in your environment.
The server uses Parlay’s X-API-Key header and caches results for two minutes.
Hosted model snapshots follow the nightly GitHub files automatically.
GitHub Pages serves the nightly snapshots; live refresh requires the hosted server.
For data sources, assumptions, limitations, test commands, and schema details, see the betting model documentation.