Day 32: Position Sizing & Kelly Criterion Analysis
Topic: Optimal position sizing for Polymarket 15-min trading
Date: 2026-03-14
Status: โ
COMPLETED
Research Question
What position sizing maximizes risk-adjusted returns?
Data Analysis
Analyzed 416 trades from trade-journal.json:
| Metric | Value |
|---|---|
| Win rate | 72.1% |
| Avg win | $10.87 |
| Avg loss | $1.87 |
| W/L ratio | 5.80 |
Kelly Criterion
- Full Kelly: 67.3% of bankroll (impractical)
- Half Kelly: 33.7% of bankroll (aggressive but survivable)
Simulated Performance by Position Size
| Position Size | Final Balance | Max Drawdown |
|---|---|---|
| 5% | $70.85 | - |
| 10% | $131.71 | 4.3% |
| 20% | $253.41 | 5.6% |
| 30% | $375.12 | 6.2% |
Key Findings
- Aggressive sizing works โ Even 20-30% of bankroll yields <6% max drawdown
- Current system uses ~$5-10 per trade โ Conservative vs Kelly suggests 10-20% is optimal
- W/L ratio is exceptional (5.8:1) โ Allows aggressive sizing
- Kelly is unrealistic โ 67% per trade is too risky; 10-20% is prudent maximum
Recommendation
NOT DEPLOYABLE โ Current sizing is conservative and appropriate. The math suggests we could size larger, but: - Black swan risk (market resoluton errors, smart contract bugs) - Execution constraints (minimum order sizes, slippage) - Psychological comfort
Current 1-2% of bankroll per trade is safe. For live trading with real money, recommend 5-10% max.
Next Steps
- Consider live trading when comfortable with position sizing
- Current dry-run validates filters; ready to transition to paper/live