Screening Cointegrated Stocks for Statistical Arbitrage
Summary
This article outlines a screening funnel for building a market-neutral mean-reversion portfolio of Nasdaq stocks cointegrated with Nvidia. It begins with a broad broker symbol universe, applies sector and industry filters to focus on companies with plausible shared drivers, and proposes scoring candidates using correlation, Engle-Granger and Johansen cointegration tests, portfolio-weight stability, ADF and KPSS spread stationarity tests, and liquidity. The purpose is to reduce a combinatorial search to candidates that are statistically promising and practically tradable.
The example uses MetaTrader symbol metadata and a database to record ticker, exchange, asset type, sector, industry, currency, and data source. Semiconductor stocks provide the initial focused group, after which statistical tests narrow the candidates. The article frames the workflow as preparation for later backtesting and demo trading, not proof of profitability. Its starting universe depends on symbols and metadata available from the chosen broker server, and liquidity, short availability, transaction costs, and changing relationships remain important constraints.
Key ideas
- Begin with a liquid, accessible universe and narrow it using sector and industry information.
- Use cointegration tests, weight stability, spread stationarity, and liquidity as screening criteria.
- A scoring funnel reduces the computational burden of testing every possible stock pair or basket.
- Store the broker or server source alongside symbol metadata so market data can be distinguished later.
- Screened candidates still require backtesting and demo trading, and must satisfy practical shorting and cost constraints.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.