Backtesting

Backtesting is a general term that refers to testing a predictive model on historical data. It is critical for regression-testing changes to models before they are promoted into production.

Let’s assume we have some trade data in our risk system that also serves as an authoritative platform for regression testing:

2024-04-01

Run

Open in xt-fiddle

The system also contains the following history of prices:

2024-04-01

Run

Open in xt-fiddle

The backtesting processes are able to simulate querying as-of successive moments in time by efficiently querying the entire database without the need for explicit snapshots, using SQL:

Run

Open in xt-fiddle

This as-of view across all database tables in the system can encompass all reference data and complete portfolio states, which thereby maximises the number of useful signals available to the model.