Edge Tracker
Whenever the model's probability beats the best market price by 5%+ expected value, a flat 1-unit paper position is opened and later settled against the real result. This measures one thing only: does model–market divergence carry signal, or does the market simply know more? No real money, no recommendations.
Ledger Summary
Entry rule: model probability × best price − 1 ≥ 5% and model probability ≥ 15% (no longshots). One position per match & outcome, locked at the first qualifying snapshot. Below ~200 settled positions the ROI is statistically meaningless — collect first, conclude later.
Settled Positions
Nothing settled yet — positions settle automatically once their match finishes.
How to read this
A sustained positive ROI over a large sample would mean the model sees something the market misprices — that pattern (not individual picks) is the asset. A negative ROI means the market is better informed than the model, and each divergence is a measure of what the model is missing (team news, lineups, money flow). Either result is valuable; only the sample size makes it trustworthy.