Prediction Market Lab

Venue-agnostic prediction market research, backtesting, and paper trading system.

Status: Backtesting complete. Ready for paper trading once terms/jurisdiction confirmed.

Architecture

┌─────────────────────────────────────────────┐
│           Command Center                    │
│  Discovery → Inventory → Regime → Scoring   │
└─────────────────────────────────────────────┘
                      ↓
┌─────────────────────────────────────────────┐
│           Safety Layer                      │
│  KillSwitch + AuditTrail (fail-closed)      │
└─────────────────────────────────────────────┘
                      ↓
┌─────────────────────────────────────────────┐
│           Venues                            │
│  Polymarket (CLOB) | Limitless (Base)       │
└─────────────────────────────────────────────┘

Backtest Results

Limitless (300 snapshots, 272 lifecycles)

Strategy	N	Hit %	Avg P&L/share
H7 [0.5-0.6]	42	67%	+0.068
H2 Late fav	132	89%	+0.023
H3 Momentum	84	86%	+0.022

Polymarket (130 wallet markets)

Metric	Value
Two-sided %	86.9%
Median pair cost	0.83 (-17%)
Pairs < 1.0	77%
Net edge (H7)	+0.066/share

Fee Comparison

Venue	Pair Cost	Arb Windows	Best For
Limitless	+8.5%	0%	Niche directional
Polymarket	-17%	77%	Momentum + Hedging

Installation

cd prediction-market-lab
python -m venv .venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows
pip install -r requirements.txt

Running Analysis

# Run hypothesis battery on Limitless data
python scripts/hypothesis_battery_full.py

# Compare venue fees
python scripts/scanner.py --compare

# Live scanner (scans Polymarket every 2 runs)
python scripts/scanner.py --runs 3

Data

• data/snapshots/ — Limitless hourly snapshots (300 files)
• data/wallet-ce25-sample.json — Polymarket wallet trade history
• scan-history.jsonl — Command center scan logs

Safety

All execution is fail-closed by default:

1. Create KILL file to halt all operations
2. All scans logged to audit/orders.jsonl
3. Execution policy requires 4 approvals

Repositories

Repo	Purpose
prediction-market-lab	This repo — backtesting, analysis, research
Predictions-Lab-tablet	Production tablet agent with execution framework

Next Steps

1. Paper trade H7 on Polymarket (small size)
2. Add real-time WebSocket feed
3. Build automated execution once legal review complete
4. Expand to Kalshi, Hyperliquid, other venues