tools
Numerai Signals
Numerai Signals lets data scientists submit stock-market predictions, measure their originality, and optionally stake NMR on performance.
Numerai Signals is a tournament for data scientists and data providers who submit numerical predictions for a global stock universe. Numerai scores signals against its targets and other submissions, provides diagnostics, and can combine staked signals in its hedge fund.
Users must bring their own market data and build their own models. Numerai provides supporting datasets and APIs, but staking uses NMR and may result in earned or burned tokens. Cloud infrastructure used with the CLI costs extra.
Features
- Submit predictions for roughly 5000 global stocks
- Compare signals against other submissions for originality
- Receive diagnostic backtests and performance metrics
- Score predictions against neutralized targets
- Download Signals v3.0 training, validation, and live datasets
- Upload predictions through SignalsAPI or the GraphQL API
- Optionally stake NMR on model performance
- Deploy automated prediction nodes with Numerai CLI
Use cases
- Submit a proprietary equity signal for out-of-sample evaluation
- Train machine-learning models on external financial data
- Test whether a signal adds information beyond existing signals
- Automate prediction uploads on a recurring schedule
- Stake NMR on a model after reviewing its diagnostics