Strategy
AI Divergence
Compare model-derived fair value with market-implied probability, then act only when the gap is large enough to be meaningful after cost, time, and uncertainty.
📌 Strategy Overview
AI Divergence looks for markets where a structured model and the live market disagree. The model may be based on polling, historical data, public indicators, or event-specific variables. The trade is not “AI is always right.” The trade is “the disagreement is large enough to investigate and potentially monetize.”
💡 Core idea: when the market is reacting slowly, emotionally, or with incomplete information, a model can provide a cleaner benchmark for fair value.
🔄 Core Logic
1. Build fair value
- Inputs: polls, macro data, historical analogs, demographics, market microstructure
- Model: Bayesian or ensemble frameworks that generate an implied event probability
- Output: a fair probability estimate, not certainty
2. Detect divergence
- Market price: the live Polymarket implied probability
- Model price: your estimated fair value
- Gap size: the absolute difference between the two
- Threshold: act only when the gap is wide enough to survive noise and costs
3. Execute with discipline
- Model > Market: buy YES if the contract looks underpriced
- Model < Market: buy NO if the contract looks overpriced
- Exit on convergence: take profit as the market reprices or new data closes the gap
⚙️ What You Need
- Data: reliable structured data, not just headlines
- Modeling workflow: a repeatable way to estimate fair value and update it
- Review layer: a human sanity check before acting on a false signal
- Execution rule: predefined thresholds for entry, sizing, and exit
📊 Advantages & Risks
Advantages
Objective framing
The model forces you to price the event instead of narrating it loosely.
Repeatability
You can review, backtest, and refine the process over time.
Cross-market use
The same logic can be adapted across politics, macro, sports, and other event markets.
Risks
- Model risk: the model can be wrong, stale, or built on bad assumptions
- Slow convergence: the market may stay irrational longer than expected
- Regime change: new events can invalidate historical relationships quickly