Quantitative research project / 03
After the Bell
Bitcoin–Equity Gap Forecasting
Research question
Do Bitcoin weekend price movements contain information about subsequent opening gaps in COIN, MSTR and MARA?
Data and design
The project uses 156 weekends of cryptocurrency and equity data. Forecasts use expanding-window ordinary least squares models, chronological evaluation and benchmark comparisons rather than a random train/test split.
Evaluation
Performance and uncertainty are examined using MAE, RMSE, directional accuracy and block-bootstrap confidence intervals. The design explicitly considers temporal dependence, information timing and alternative explanations.
Reproducibility
A Python and SQL pipeline includes validation tests for market calendars, daylight-saving transitions, target construction and data leakage. Code and an interactive historical replay are linked below.
Interpretation
The project examines predictive association in historical data. It does not establish causality, and its historical forecasting comparison is not evidence of an implementable or profitable trading strategy.