Validated telemetry — not marketing claims.
Self-checking surfaces over pipeline tables: calibration, coverage, integrity sentinels, and live hedge-alert precision.
Per-model isotonic maps fitted out-of-sample; regression guard only applies maps that beat raw on holdout Brier.
Every alertable hedge window is written append-only and scored post-horizon — precision is measurable, not asserted.
Movement-selected ~150-market universe with disclosed capture rate (~one-fifth of top movers on a typical day).
Same signal content and thresholds for every entitled customer. Tiers gate delivery mechanics only.
Free proof surfaces on the public API service — embargoed stress index (24h) and scored alert ledger (7d). Linked from Track record.
GET /api/v1/risk/status returns per-table freshness vs SLA, active-pin coverage,
trailing live hedge-alert precision (overall + per arm), and calibration status including out-of-fit verdicts.
Enterprise customers also get /api/v1/enterprise/status for feed freshness.
Universe coverage (~22% of top movers on a given day). Signals describe the tracked universe, not the entire venue.
Stress and clusters detect broad repricing as it happens — validated as coincident measurement, not leading forecasts.
Shadow posture: nothing executes. All outputs are decision support.
“We measure whether the system is producing quality data — and publish when it isn't.”
Standardized, impersonal market analytics for informational purposes only — not trading, investment, or hedging advice, and not a recommendation to transact in any market. Outputs are identical for all customers and are not tailored to any person's positions or circumstances (we never ingest customer positions, inventory, or P&L). Past performance, including live scored history and backtests, does not guarantee future results. All decisions remain yours.