Proxy
An independent shadow model of a fund's daily performance from its holdings.
Proxy is an independent shadow model that predicts what a fund's daily performance should be, built position by position from the portfolio's actual holdings. It flags when a fund's reported NAV moves in a way its holdings don't justify. It answers: does this fund's reported return actually match what its holdings did today?
How it works
It builds an independent, position-by-position prediction of what the portfolio should earn each day and compares it against the administrator's reported NAV. For each holding it uses an enriched representation (adding fields such as notional exposure, implied volatility, and delta) and calibrates an instrument-specific pricing model, falling back to a benchmark-based estimate when a position lacks data, then aggregates all the contributions plus FX, dividend, and coupon effects into one predicted daily profit and loss. It tracks the daily gap (the tracking error) against the reported value and flags an anomaly when that gap is abnormal versus the fund's recent history, using a method that adapts to each strategy's normal volatility rather than a fixed threshold.
What you get
- A daily predicted profit and loss, built bottom-up from every position and adjusted for FX, dividends, and coupons.
- Position-level metrics not usually in fund-admin data, such as notional exposure, implied volatility, and delta.
- A daily tracking error between the prediction and the reported value.
- Anomaly flags when that tracking error is abnormal versus the fund's own recent history.
How to use it in Spark
The predicted returns and tracking error arrive as data in your workspace. Chart predicted versus reported on a canvas, and let a workflow or the Agent act when the tracking error looks anomalous.