Research Platform

Five-day forecasts, measured honestly

Gilt & Willow is a study of how well — and how poorly — statistical models predict short-term movements in the S&P 500. The goal is not a profitable signal. It is a reproducible, well-measured answer to the question of whether a model beats doing nothing at all.

A five-day horizon

The platform forecasts the S&P 500's return over the next five trading days — not calendar days, so weekends and market holidays are excluded.

Baseline first

Every model is reported next to a naive baseline that simply predicts no movement. A model that cannot beat that baseline has not earned any confidence.

Honest measurement

Each forecast is scored against what actually happened once the five days have elapsed. Predictions are recorded before the fact and never revised.

Current status

Phase 1 is in progress. The web application and the Python forecasting service are scaffolded and running; authentication, market-data ingestion and the forecasting models themselves are not yet built. No forecast is available from this deployment yet.

Web health check
/api/health
Forecasting service health check
localhost:8000/health