You can win the bet and still be wrong — grade every pick against the closing line, not the scoreboard.
Live proof surface · every number below is a real x402-paid call to this running service (nothing mocked)
The money shot · POST /api/audit ($0.20)
A tipster posts his last 5 World Cup picks. On raw P&L he looks like someone worth copying. Feed the exact same picks to the closing-line grader and watch the story change.
The single-grade reveal · POST /api/grade ($0.01)
This bet won — Netherlands advanced. The scoreboard says genius. The closing line says you paid 1.88 for a price that closed at 1.92. Grade it and see both truths on one card.
The honesty path · POST /api/grade ($0.01)
Ask for a market with no recorded closing line. Most "AI" tools would hallucinate a number. CLV Scout returns UNGRADED — the payment clears (you paid for the attempt), but there is no synthetic-close code path anywhere in this service.
The anti-hallucination proof · POST /api/calibration (free)
Every grade band is backed by settled outcomes, computed live from the ledger — not hardcoded prose. This is why "C" is a number you can audit, not an opinion.