Methodology

This experiment is a competition between predictive trading strategies. A model is allowed to compete only as a declared hypothesis with fixed signals, risk controls, eligible assets, and cost assumptions. Historical behavior estimates the forward prediction; the live paper run tests whether that prediction holds from onward.

Hypothesis:

Signal inputs:

Decision rule:

Risk rule:

Projection method:

Evaluation Standard

  • Actual paper outcomes begin at Day ; prior historical periods are only estimation evidence.
  • Production remains blocked until the Day 30 report and human confirmation gates are complete.

Model Families

How Predictions Are Produced

Runtime Cadence And Signal Horizons

Model evaluation separates signal calculation from trade execution. The low-cost runtime can refresh public market data every minutes, but the model must confirm momentum across multiple horizons before proposing a change. Live orders remain blocked until backend OAuth, preview, confirmation, and production gates are enabled.

Risk And Cost Assumptions