Trade Journal

Systematic trade logging and pattern detection for Polymarket.

Trade Logging

Every trade gets logged to ~/obsidian-vault/memory/trade-journal/YYYY-MM-DD.json:

[
  {
    "timestamp": "2026-03-26T14:30:00Z",
    "asset": "BTC > 90k by April",
    "market_id": "0x...",
    "strategy": "FairValueGap",
    "timeframe": "1d",
    "entry_price": 0.45,
    "exit_price": 0.52,
    "signal_type": "reversal",
    "conviction": "high",
    "size_usd": 5.00,
    "outcome": "W",
    "pnl_usd": 0.35,
    "fees_usd": 0.10,
    "notes": "Clean signal, good sizing"
  }
]

Fields

When to Log

• Entry: Log immediately with outcome "open"
• Exit: Update entry with exit_price, outcome, pnl_usd, fees_usd
• End of day: Verify all trades logged, fill gaps from bot logs

Weekly Summary

Generate ~/obsidian-vault/memory/trade-journal/weekly-YYYY-WXX.md every Sunday:

# Trade Journal — Week XX, YYYY

## Stats
- Total trades: X
- Win rate: X%
- Total PnL: $X.XX (after fees: $X.XX)
- Avg win: $X.XX | Avg loss: $X.XX
- Profit factor: X.XX

## WR by Strategy
- FairValueGap: X% (N trades)
- MeanReversion: X% (N trades)

## WR by Signal Type
- reversal: X% (N trades)
- momentum: X% (N trades)

## WR by Conviction
- high: X% (N trades)
- medium: X% (N trades)

## WR by Asset (top 5)
- ...

## Patterns Detected
- ...

## Sizing Analysis
- Avg size on wins: $X.XX
- Avg size on losses: $X.XX
- Conviction-size alignment: [good/poor]

Pattern Detection Rules (flag automatically)

• Any asset with WR < 40% over 5+ trades -> "Stop trading X on Y timeframe"
• High conviction WR < 50% -> "Conviction calibration is off"
• Avg win size < avg loss size -> "Sizing is inverted — cut losers faster or size winners bigger"
• Signal type with WR > 65% over 10+ trades -> "Double down on X signals"
• Strategy with negative PnL after fees over 20+ trades -> "Kill this strategy"
• Fee-to-PnL ratio > 50% -> "Fees are eating your edge"

Pulling Data from Polymarket API

POLY_ADDRESS="<bot-wallet-address>"
API_URL="https://clob.polymarket.com"
DATE=$(date +%Y-%m-%d)
OUTPUT="memory/trade-journal/${DATE}.json"

curl -s "${API_URL}/trades?maker=${POLY_ADDRESS}&limit=100" | \
  python3 -c "
import json, sys
trades = json.load(sys.stdin)
formatted = []
for t in trades:
    formatted.append({
        'timestamp': t.get('timestamp',''),
        'asset': t.get('market',''),
        'market_id': t.get('asset_id',''),
        'entry_price': float(t.get('price',0)),
        'size_usd': float(t.get('size',0)),
        'signal_type': 'bot',
        'conviction': 'auto',
        'outcome': 'open',
        'notes': ''
    })
print(json.dumps(formatted, indent=2))
" > "$OUTPUT"

Lessons (from real losses)

• Low-cent positions across the board — even high-conviction got low fills. Minimum $0.50 on high conviction.
• Log conviction at signal time, not retroactively. Bot should tag each trade.
• Fees destroy marginal edges — only strategies with 70%+ WR survive 2% maker fees.