【广发金工】用逐笔订单数据改进分钟频因子：海量Level 2数据因子挖掘系列（六）

Core Insights - The article emphasizes the importance of data collection and analysis for quantitative investors to uncover hidden market patterns and gain an edge in stock market trading [1][4][5]. Group 1: Data Types and Importance - Level 1 data includes basic market information such as highest price, lowest price, opening price, closing price, trading volume, and trading amount, updated every three seconds [6][7]. - Level 2 data provides more detailed information, including tick data that captures every order during trading sessions, allowing for deeper analysis of market trends and trading signals [6][9]. Group 2: Key Period Factors - The article introduces a set of Level 2 factors based on key trading periods, categorized into four main types: price changes, price levels, trading amounts, and volume-price coordination, totaling 123 factors [12]. - Specific factors such as KeyPeriod_ret_zero and KeyPeriod_ret_low5pct show historical RankIC averages of -5.36% and 5.47% respectively, with win rates of 85.1% and 84.1% [2]. Group 3: Factor Performance - The performance of various factors is highlighted, with low price period factors like KeyPeriod_price_low5pct achieving a 20-day RankIC average of 5.59% and a win rate of 85.3% [2]. - Trading amount factors such as KeyPeriod_amount_top30pct show a 20-day RankIC average of 11.23% with a win rate of 84.8%, indicating strong predictive power [2]. Group 4: Research and Development - The article outlines ongoing research efforts to refine and develop new factors from Level 2 data, with a focus on enhancing the predictive capabilities of trading strategies [10][12]. - Previous reports have successfully identified effective factors, with some achieving historical RankIC averages above 9.2% and win rates around 76% [10].