v2.1 Dataset — Numerai Signals
The "Alpha" dataset was released in July 2025.
- train.parquet
- Data used to train your model
- train_neutralizer.parquet
- Neutralization matrix for training eras. Signals Alpha and MPC use this matrix to neutralize your signal. Useful for training a model that is neutralization-aware.
- train_sample_weights.parquet
- Sample weights vector for training eras. Signals Alpha and MPC use sample weights to weight your signal after neutralization. Useful for training a model that is sample-weight-aware.
- validation.parquet
- Data used to validate or train your model.Expands every week to include data for the most recent eras with targets available.
- validation_neutralizer.parquet
- Neutralization matrix for validation eras. Signals Alpha and MPC use neutralize your signal using this matrix. Useful for calculating Alpha over the validation set.
- validation_sample_weights.parquet
- Sample weights vector for validation eras. Signals Alpha and MPC use sample weights to weight your signal after neutralization. Useful for calculating Alpha over the validation set.
- validation_example_preds.parquet
- An example file with predictions on the validation.parquet file. Allows you to easily try diagnostics.
- live.parquet
- The live data that your model predicts on. This changes daily.
- live_neutralizer.parquet
- Neutralization matrix for live data. Signals Alpha and MPC use this matrix to neutralize your signal.
- live_sample_weights.parquet
- Sample weights vector for live data. Signals Alpha and MPC use these weights to weight your signal after neutralization.
- live_example_preds.parquet
- An example file with predictions on the live.parquet file. Allows you to easily try your first submission.