Complete decision ledger
Forecasts and executable prices are frozen at decision time
No AI verdicts yetThe ledger populates only after a real AI research scan.
SYSTEM PLAYBOOK · PAPER ONLY
Three horizons. Three different edges.
The lab is deliberately opportunity-seeking, but every paper position must pass horizon-specific market quality, AI research, confidence and edge rules. The 15-minute watcher never spends AI tokens.
FAST · 1–7 DAYSCatalyst Sprint
Targets imminent, verifiable catalysts where public information may reprice quickly. Wider market tolerance, but demands the largest AI edge.
- Liquidity ≥ $50k
- Spread ≤ 8 pts
- Entry edge ≥ 10 pts
- Confidence width ≤ 30 pts
- Stake $250
- Max new positions 2
MEDIUM · 8–30 DAYSCatalyst + Trend
Balances fresh evidence, base rates and the next dated catalyst. Designed as the core strategy with moderate turnover.
- Liquidity ≥ $100k
- Spread ≤ 6 pts
- Entry edge ≥ 8 pts
- Confidence width ≤ 25 pts
- Stake $200
- Max new positions 2
LONG · 30+ DAYSStructural Mispricing
Looks for durable disagreement with the market. Liquidity and calibration are stricter because capital stays committed longer.
- Liquidity ≥ $250k
- Spread ≤ 5 pts
- Entry edge ≥ 6 pts
- Confidence width ≤ 20 pts
- Stake $150
- Max new positions 2
01Mathematical shortlist
Load active Polymarket markets, then reject weak liquidity, wide spreads, extreme prices, duplicate positions and games that have started.
02AI web research
Read the exact resolution rules, search current primary sources and estimate YES probability independently from the market.
03Edge calculation
YES edge = AI probability − executable YES price. NO edge = (1 − AI probability) − executable NO price.
04Paper entry & settlement
Apply the selected strategy’s stake and position cap. When Polymarket resolves, calculate realized P&L and preserve the audit trail.
Rules shared by all three
- Executable CLOB prices required
- Existing open market rejected
- Started sports market rejected
- Sports safety margin 30 minutes
- Primary sources and rules required
- Ambiguous resolution SKIP
- OpenAI runs manual only
What each verdict means
- ENTER YES — YES passed the edge threshold
- ENTER NO — NO passed the edge threshold
- WATCH — researched, but edge is insufficient
- SKIP — ambiguous, stale or poorly sourced
- VOID — position invalidated without P&L
Important: edge is an estimate, not guaranteed profit. The lab remains paper-only until enough positions resolve to evaluate calibration, Brier score and realized returns.