Class-oriented TypeScript scanner for soccer betting edges — pull feeds, price markets, replay portfolio history, and emit live signals. Uses feed / book / vault CLI vocabulary with renamed domain types.

• Feeds — SampleLeagueFeed (in-memory) and GithubSoccerFeed (cached remote CSV)
• Pricing — MarketEdgePricer compares mean implied probabilities against posted odds
• Portfolio — backtestRunner with TemporalCv walk-forward splits
• Vault — Redis-backed cache layer with memory fallback
• Distinct naming — AbstractFeed, BasePricer, TemporalCv (not toolbox names)

Requirements: Node.js 20+

npm install
npm test
npm run scan -- feed schema
npm run scan -- book signals --no-cache

soccer-edge-market-scanner/
├── scanner/run.ts              CLI entry
├── core/                       schema.ts, frameKit.ts, checks.ts
├── feeds/                      AbstractFeed, SampleLeagueFeed, GithubSoccerFeed
├── pricing/                    BasePricer, MarketEdgePricer, ModelPricer
├── portfolio/                  backtestRunner, TemporalCv, metrics.ts
├── storage/                    redisConn, cacheLayer, cacheKeys
├── src/index.ts                Public library barrel
├── __tests__/                  Vitest tests
└── playbook/                   Scan workflow docs

flowchart LR
  F[feed pull] --> P[SampleLeagueFeed]
  P --> B[book replay]
  B --> M[MarketEdgePricer]
  M --> R[Portfolio metrics]
  P --> S[book signals]
  S --> V[vault cache]

npm run scan -- feed schema
npm run scan -- feed pull --out ./feeds
npm run scan -- book replay --out ./replay
npm run scan -- book signals
npm run scan -- vault ping

import {
  SampleLeagueFeed,
  MarketEdgePricer,
  backtest,
  TemporalCv,
} from "./src/index.js";

const feed = new SampleLeagueFeed();
const [X, Y, O] = feed.extractTrainData(0, "williamhill");

const pricer = new MarketEdgePricer(["williamhill"], 0.03);
pricer.fit(X, Y);

if (O) {
  const replay = backtest(pricer, X, Y, O, new TemporalCv(2));
  console.log(replay);
}

• playbook/scan-workflow.md — feed → book → vault flow