高频因子跟踪

• The report tracks high-frequency stock selection factors, including price range factor, price-volume divergence factor, regret avoidance factor, and slope convexity factor, with their out-of-sample performance being generally strong[2][3][11] - Price Range Factor: Measures the activity of stock transactions within different intraday price ranges, reflecting investors' expectations of future stock trends. High price range transaction volume and transaction count factors are negatively correlated with future stock returns, while low price range average transaction volume factor is positively correlated with future stock returns. The factor is constructed by combining three sub-factors: high price 80% range transaction volume factor (VH80TAW), high price 80% range transaction count factor (MIH80TAW), and low price 10% range average transaction volume factor (VPML10TAW). These sub-factors are weighted at 25%, 25%, and 50%, respectively, and are industry market value neutralized[12][14][17] - Price-Volume Divergence Factor: Measures the correlation between stock price and trading volume. When price and volume diverge, the likelihood of future price increases is higher, while convergence indicates a higher likelihood of price decreases. The factor is constructed using high-frequency snapshot data to calculate the correlation between snapshot transaction price and snapshot trading volume, as well as snapshot transaction price and transaction count. Two sub-factors are used: price and transaction count correlation factor (CorrPM) and price and trading volume correlation factor (CorrPV). These sub-factors are equally weighted and industry market value neutralized[22][23][25] - Regret Avoidance Factor: Based on behavioral finance theory, this factor utilizes investors' regret avoidance emotions to construct effective stock selection factors. It examines the proportion and degree of stock price rebound after being sold by investors. The factor is constructed using transaction data to identify active buy/sell directions, with additional restrictions on small orders and closing trades to enhance performance. Two sub-factors are used: sell rebound proportion factor (LCVOLESW) and sell rebound deviation factor (LCPESW). These sub-factors are equally weighted and industry market value neutralized[26][32][35] - Slope Convexity Factor: Derived from the elasticity of supply and demand, this factor uses high-frequency snapshot data from limit order books to calculate the slope and convexity of buy and sell orders. The factor is constructed by aggregating order volume data by level and calculating the slope of buy and sell order books. Two sub-factors are used: low-level slope factor (Slope_abl) and high-level seller convexity factor (Slope_alh). These sub-factors are equally weighted and industry market value neutralized[36][41][43] - High-frequency "Gold" Portfolio Strategy: Combines the three high-frequency factors (price range, price-volume divergence, and regret avoidance) with equal weights to construct an enhanced strategy for the CSI 1000 Index. The strategy includes mechanisms to reduce transaction costs, such as weekly rebalancing and turnover rate buffering. The strategy's annualized excess return is 10.20%, with an IR of 2.38 and maximum excess drawdown of 6.04%[44][46][47] - High-frequency & Fundamental Resonance Portfolio Strategy: Combines high-frequency factors with fundamental factors (consensus expectations, growth, and technical factors) to construct an enhanced strategy for the CSI 1000 Index. The strategy's annualized excess return is 14.49%, with an IR of 3.46 and maximum excess drawdown of 4.52%[48][50][52]