Alpha Signal Discovery Challenge
Discover novel alpha signals from obfuscated market data. Participants must build predictive models that generate uncorrelated returns across multiple market regimes. The dataset spans 5 years of multi-asset returns with engineered features — column names and asset identities are hidden to prevent data leakage.
Prize Pool
$25,000
Participants
234
Timeline
Feb 1 — Mar 31, 2026
Discover novel alpha signals from obfuscated market data. Participants must build predictive models that generate uncorrelated returns across multiple market regimes. The dataset spans 5 years of multi-asset returns with engineered features — column names and asset identities are hidden to prevent data leakage.
Evaluation
Models are scored on annualized Sharpe Ratio computed over a held-out out-of-sample test period. An overfitting penalty is applied when the divergence between in-sample and out-of-sample performance exceeds 20%. Final rankings are determined after the competition closes using a separate, hidden evaluation dataset.
Columns
147
Obfuscated features
Rows
325,000
Training observations
Scoring
Sharpe Ratio (annualized)
Primary evaluation metric
Runtime Limit
30 minutes per evaluation
Memory Limit
8GB RAM
Max Submissions
100 total