jev-regime-gate

Jev-powered regime gate for trading strategy backtests. Uses TypeSafe's Jev decision model to classify market regimes on daily bars, then gates a strategy by regime compatibility and confidence.

Research experiment, not investment advice.

Pipeline

1. Encode — daily OHLCV → pre-bucketed feature summary (no raw prices to Jev)
2. Classify — one POST to Jev with three typed questions: regime_type, regime_change_likely, strategy_viable
3. Gate — trade / stand down / scale size based on regime fit + confidence
4. Compare — same strategy, same period: gated vs ungated equity curves

Results are stored per bar in OpenCandle meta fields.

Usage

cp .env.example .env  # add your OpenRouter API key
npm install
npm run fetch-data    # pull daily bars
npm start             # run backtest

Schema

Each bar's meta.jev contains:

{
  "regime_type": "trend_up",
  "regime_max_p": 0.82,
  "regime_change_likely": 0.23,
  "strategy_viable": 0.91,
  "gate_decision": "trade",
  "size_factor": 1.0
}

License

MIT — Matchstick Trading