Low-Frequency Lead/Lag Screening for Cointegrated Assets
Summary
The article explains how to search for slower lead/follower behavior between related assets, where one instrument’s returns may precede another’s by minutes or longer. It frames the setup as a potential mean-reversion opportunity for cointegrated assets and contrasts it with higher-risk momentum ideas for assets without an established relationship. Cross-correlation across return series at different time offsets is the core screening method; a synthetic example illustrates how a known delay should appear as a correlation peak at the corresponding lag.
The workflow combines a Python analyzer, ordered SQL window calculations, and a MetaTrader 5 backtest. It emphasizes deterministic row numbering, explicit time ordering, and timestamp alignment to avoid misleading results. The article reports that a sample pair without a non-zero-lag peak did not look tradeable, then describes testing another pair and optimizing rolling windows and an exit threshold. This is an exploratory case study, not evidence of a robust edge: the excerpt gives no detailed performance statistics, and backtest findings do not establish future profitability.
Key ideas
- Cross-correlation of returns at shifted times can identify candidate lead/follower relationships.
- A non-zero-lag correlation peak is a screening signal, not proof that the delay can be traded profitably.
- Cointegrated pairs may support a mean-reversion interpretation when one asset temporarily lags the other.
- SQL window calculations need deterministic ordering and aligned timestamps to produce reproducible results.
- Backtesting and parameter selection are needed to assess candidate relationships, with overfitting remaining a concern.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.