Key points of the report
This paper tests the single factor performance of Barra style factor system and algorithm factor system in the last week. Last week, the style switching accelerated, the turnover factor increased significantly, showing the best negative selection effect, and the scale retreated. In addition to the turnover rate, the performance of volatility and long-term reversal factor has also been enhanced. The exposure can be appropriately increased in low turnover and low volatility style, and the whole remains balanced.
Summary:
Single factor performance in the last week (industry wide): among the Barra style factor system, monthlyshare turnover (monthly turnover rate), quarterly turnover (quarterly turnover rate) and longterm relative strength (long-term relative strength) performed best. In the algorithm mining / machine learning factor system, alpha006 (trend class), alpha032 (trend class) and alpha046 (inversion class) perform best.
Factor performance of CITIC’s primary industry: according to the classification method of CITIC’s primary industry, it is divided into industries. In each subdivided primary industry, the daily frequency data of the last week is used to calculate and display the expression of single factor in Barra style factor system and algorithm mining / machine learning factor system. In the coal industry with the best performance, sizelncap (market value scale) and MIDCAP (medium market value) in Barra system perform best (according to rankic mean and rankiir), and alpha023 (inversion class) and alpha049 (inversion class) in algorithm system perform best (according to rankic mean and rankiir); Among the worst performing consumer service industries, longterm relativestrength (long-term historical alpha) and MIDCAP (medium market value) in Barra system perform best (according to rankic mean and rankiir respectively), and alpha009 (inversion class) and alpha101 (inversion class) in algorithm system perform best (according to rankic mean and rankiir respectively).
Performance of single factor in time series: calculate and display the performance of single factor in time series according to style categories using the data of the last half and the last three respectively. In the last week, the multi factor rankic has changed significantly and the factor rotation has accelerated. In the short term, it can continue to maintain the dispersion of the combination and appropriately add a better style.
