Algorithmic Sister Stocks for Pairs Trading
Автор: Eugene A.
Загружено: 2025-12-01
Просмотров: 5
This document details the multi-phase implementation plan for a high-speed Python system dedicated to discovering optimal pairs for statistical arbitrage from 10,000 5-minute stock data files. The initial phases focus on cleaning data, efficiently handling liquidity filters, and aligning time series using parallel loading to prepare the large dataset for analysis. The core engine applies advanced statistical metrics, including Engle-Granger cointegration tests, Ordinary Least Squares regression to determine hedge ratios, and mean reversion calculations using the Half-Life and Hurst Exponent. A crucial efficiency step involves a preliminary multi-timeframe correlation filter to quickly screen millions of potential pairs before running expensive statistical tests. Performance is heavily optimized through the use of parallel processing across 36 cores, aiming to complete the entire analysis in under 30 minutes. All results are synthesized into a Composite Score for ranking, generating detailed CSV output and complex diagnostic visualizations for the highest quality pairs.
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке:
