Sub-Fund Evolution
Predict how a fund should behave day to day and flag deviations.
Sub-Fund Evolution predicts how a fund's portfolio should behave day to day, then flags when the actual behavior deviates. It combines several independent models to catch portfolio anomalies and help pinpoint their source. It answers: is my fund moving the way it should, or is something anomalous happening?
How it works
It builds an independent statistical prediction of how the portfolio should behave, then flags when actual behavior deviates. It combines three models: a synthetic proxy that estimates each position's profit and loss, a factor-based nowcasting model that explains returns through market benchmarks and known factors, and a peers model that clusters similar portfolios. The three signals are merged into one, weighted by each model's confidence score, and an anomaly is raised when the combined signal crosses its threshold.
What you get
- A predicted expected return for the portfolio, built by combining independent models.
- A synthetic proxy that mirrors the portfolio by estimating each position's profit and loss.
- A peer-comparison view that clusters similar portfolios to spot movements that stand out from the group.
- An anomaly event that fires when the combined signal breaches its threshold.
How to use it in Spark
The predictions and anomaly events arrive as data in your workspace. Track expected versus actual on a canvas, and when an anomaly fires, let a workflow or the Agent raise a task to investigate.