Polymarket Optimizer

Automatic parameter optimizer for polymarket-executor. Reads performance_metrics.json every 6 hours, analyzes win rates and P&L per strategy, adjusts learned_config.json to improve future performance. Also builds paper trade metrics and assesses live trading readiness. Part of the Wesley Agent Ecosystem — mirrors crypto-executor-optimizer pattern.

Polymarket Optimizer — Adaptive Parameter Tuner

🧠 WHAT IT DOES

Reads trading performance from polymarket-executor, analyzes what works, and adjusts parameters automatically.

Runs every 6 hours via OpenClaw cron job. Zero manual intervention required.

Input files (written by executor):

• performance_metrics.json — per-strategy win rates, P&L, trade counts
• paper_trades.json — all simulated trade records
• portfolio.json — capital, positions, daily P&L

Output files (read by executor at next startup):

• learned_config.json — updated thresholds, Kelly fraction, allocations
• optimizer_log.jsonl — full history of every optimization run

---

⚡ QUICK START

bash

cd /data/.openclaw/workspace/skills/polymarket-optimizer
python3 polymarket_optimizer.py

Expected output:

text

============================================================
POLYMARKET OPTIMIZER v1.0.0
Run time: 2026-03-05 14:00:00 UTC
============================================================
[CONFIG] Loaded. Optimization #1
[PORTFOLIO] Health: HEALTHY
[PORTFOLIO] Capital: $102.34 | Return: 2.3%
[ANALYSIS] parity_arbitrage: EXCELLENT | WR=100.0% | Trades=8 | P&L=+0.523
[ANALYSIS] tail_end: GOOD | WR=75.0% | Trades=12 | P&L=+0.312
[OPTIMIZER] 2 adjustments made:
  🚀 parity_arb excellent → min_profit 0.020 → 0.015
  📈 parity_arb allocation 30% → 40%
[READINESS] Resolved trades: 8/30
[READINESS] Win rate: 84.6% (need 55%+)
[TELEGRAM] Report sent.
============================================================
OPTIMIZER COMPLETE
============================================================

---

🔧 WHAT IT ADJUSTS

Strategy Thresholds

If strategy is...	Action
Poor (WR < 50%)	Raise threshold → trade only best opportunities
Underperforming (WR 50–65%)	Slight threshold increase
Average (WR 65–80%)	No change
Good (WR 65–80%, positive P&L)	No change
Excellent (WR > 80%)	Lower threshold → capture more volume

Capital Allocation

If strategy is...	Action
Poor	Reduce allocation (min 5–10%)
Excellent	Increase allocation (max 50–70%)

Allocations always normalized to sum to 100%.

Kelly Fraction

Portfolio health	Action
Critical (return < -10%)	Reduce Kelly → smaller positions
Excellent (return > 10%, WR > 65%)	Increase Kelly → larger positions

Scan Frequency

Condition	Action
WR > 80% + 20+ trades	Scan faster (min 120s)
Portfolio warning/critical	Scan slower (max 600s)

---

📊 LIVE READINESS ASSESSMENT

Every run evaluates whether paper trading results justify going live.

4 criteria must ALL pass:

Criterion	Threshold
Resolved trades	≥ 30
Win rate	≥ 55%
Total P&L	Positive
Circuit breaker	Not active

Telegram report includes readiness status automatically.

---

📱 TELEGRAM REPORT

text

🧠 POLYMARKET OPTIMIZER REPORT
🕐 2026-03-05 14:00 UTC
🔄 Optimization #4

Portfolio Health: 🟢 HEALTHY
💰 Capital: $108.45
📈 Return: 8.5%
🎯 Win Rate: 76.9%
📊 Total Trades: 20

Strategy Performance:
  🚀 parity_arbitrage: 100.0% WR | 8 trades | P&L: +0.523
  ✅ tail_end: 75.0% WR | 12 trades | P&L: +0.312

Adjustments Made:
  🚀 parity_arb excellent → min_profit 0.020 → 0.015
  📈 parity_arb allocation 30% → 40%
  ⚖️ Allocations renormalized

Live Trading Readiness:
🔴 NOT YET READY
  ❌ Need 30+ resolved trades
  📊 Progress: 20/30 trades
  🎯 Win rate: 76.9% (need 55%+) ✅

---

📁 FILES

File	Location	Description
polymarket_optimizer.py	skills/polymarket-optimizer/	Main optimizer script
learned_config.json	WORKSPACE/	Output — read by executor
optimizer_log.jsonl	WORKSPACE/	Full optimization history

---

⏰ CRON JOB

Add to OpenClaw cron configuration:

text

# Run optimizer every 6 hours
0 */6 * * * docker exec openclaw-yyvg-openclaw-1 python3 /data/.openclaw/workspace/skills/polymarket-optimizer/polymarket_optimizer.py

---

Version: 1.0.0 | License: MIT | Author: Georges Andronescu (Wesley Armando)