Funding rate mean reversion trading system for Boros (Pendle's funding rate platform). Captures alpha when funding rates deviate from historical norms - rates that spike tend to revert, rates that crash tend to recover. Backtested +9% over 90 days on HYPE with 89% win rate.

Key insight: Original backtest assumed we lock in floating rate. In reality, Boros implied rate already prices in some reversion. Realistic expectation: ~22% APY.

Hold for 7 days. Reversion happens fast (2-6h) but profit accumulates AFTER.

pnpm install
cp .env.example .env

1. Message @BotFather on Telegram, send /newbot
2. Copy the bot token
3. Message your new bot (send "hello")
4. Get chat ID: Visit https://api.telegram.org/bot<TOKEN>/getUpdates
5. Add to .env:

   TELEGRAM_BOT_TOKEN=your_bot_token
   TELEGRAM_CHAT_ID=your_chat_id

# Strategy: mean_reversion (recommended) | spread_harvest | both
ALERT_STRATEGY=mean_reversion

# Hold period (7 days - don't exit early)
ALERT_HOLD_DAYS=7

# Default thresholds
ALERT_Z_THRESHOLD=2.2
ALERT_MIN_SPREAD=0.05

# Per-coin Z thresholds (optimized from backtest)
ALERT_Z_THRESHOLD_HYPE=2.5   # Stronger edge, higher threshold
ALERT_Z_THRESHOLD_BTC=2.2
ALERT_Z_THRESHOLD_ETH=1.8    # More signals, same win rate

# Coins
ALERT_COINS=HYPE,BTC,ETH

# 1. Fetch historical data (required for z-score baseline)
pnpm run data:fetch

# 2. Run backtest (optional - verify strategy performance)
pnpm run backtest

# 3. Start alerts
pnpm run alerts              # Single check
pnpm run alerts:watch        # Continuous (foreground)
pnpm run alerts:start        # PM2 daemon (auto-restart)

pnpm run alerts:start        # Start daemon
pnpm run alerts:stop         # Stop daemon
pnpm run alerts:restart      # Restart daemon
pnpm run alerts:logs         # View logs
pnpm run alerts:status       # Check status

To survive server reboots:

pm2 startup                  # Generate startup script
pm2 save                     # Save current processes

Historical data for z-score calculations is refreshed daily via cron:

# Add to crontab (crontab -e)
0 6 * * * cd /path/to/meridian && pnpm run data:fetch >> logs/data-fetch.log 2>&1

This keeps the 90-day baseline stats current. The monitor sends a Telegram notification on startup to confirm it's running.

Enter when funding deviates significantly from historical mean.

Per-coin thresholds (optimized):

Hold for 7 days. Don't exit when z-score reverts - profit comes from accumulating spread over time.

Note: Spread harvest is paused during paper trading. Mean reversion shows 6.7x better returns. May revisit later.

Enter when Boros implied rate diverges from underlying. Hold 14 days.

src/
  data/
    fetch.ts          # Fetch from Hyperliquid + Boros
    analyze.ts        # Stats and signal analysis
  backtest/
    run.ts            # Mean reversion backtest
    sensitivity.ts    # Implied rate model testing
    spread-harvest.ts # Spread harvest backtest
  alerts/
    monitor.ts        # Rate monitoring + alerts
    notifiers.ts      # Telegram + email
    config.ts         # Configuration
  paper/
    tracker.ts        # Trade lifecycle management
    monitor.ts        # Position monitoring (exit checks)
    dashboard.ts      # Full metrics display
    cli.ts            # CLI commands

From sensitivity analysis - when to use leverage:

With 2x leverage on z 2.5-4.0: +53% more PnL ($844 vs $551 in backtest).

Leverage & Liquidation:

• Boros positions use leverage - you CAN be liquidated
• Funding rates can spike unexpectedly (100%+ APR moves happen)
• A 7-day hold with adverse funding can drain margin quickly
• Start with 1-2x leverage max until you understand the mechanics
• Never use leverage you can't afford to lose

General:

• Backtest ≠ live performance
• Only 90 days of data, 9 trades - statistically limited
• Edge may compress as more capital enters Boros
• Use appropriate position sizing (10-20% of allocated capital)
• This is NOT financial advice

MIT - @0xSmartCrypto

• X - @0xSmartCrypto