Misprice

A mispricing engine, not a dashboard.

Misprice is built as a data pipeline with a single goal: turn noisy market prices into a clean, comparable truth set—then surface edges you can act on with confidence.

The core loop (one cycle)

  1. Pull prices from multiple sources
  2. Normalize into canonical events + outcomes
  3. Store snapshots with freshness + source health
  4. Estimate fair probabilities (truth model)
  5. Detect edges (+EV / arb / divergence)
  6. Log results with evidence hashes

What the UI shows

  • System mode: LIVE / DEGRADED / MOCK
  • Adapter health per source
  • Freshness of each snapshot
  • Ranked edges with EV, liquidity, confidence, and reasoning

Why this is different

Most tools only compare odds. Misprice compares markets—sportsbooks vs prediction markets—and keeps an auditable record of what the system knew at the time the signal was created.

Risk controls (kill switches)

Edges can be blocked or downgraded automatically when:

  • data is stale
  • liquidity is too thin
  • overround is too high
  • volatility is abnormal
  • confidence is below threshold
  • event start time is too close

If you want receipts instead of opinions, request access.