量化可转债研究之十三：可转债组合的风险中性方法对比

• The report discusses three risk-neutral methods for convertible bond portfolios: stratified sampling, regression residuals, and optimizer methods[6][26][39] - Stratified sampling method involves matching benchmark industry weights, segmenting by market value within industries, and selecting bonds based on quantitative factor scores[27] - Regression residuals method involves linear regression to remove market value and industry effects from alpha factors, using residuals to select bonds[28][30][31] - Optimizer method incorporates risk-neutral constraints into a portfolio optimization framework to maximize risk-adjusted returns, using numerical optimization algorithms[32][33][34][35] - Stratified sampling method shows an annualized return of 13.9%, annualized volatility of 11.4%, and a maximum drawdown of -17.2%[6][45][48] - Regression residuals method shows an annualized return of 10.9%, annualized volatility of 11.9%, and a maximum drawdown of -17.1%[6][53] - Optimizer method shows the highest annualized return of 23.1%, annualized volatility of 16.7%, and a maximum drawdown of -18.4%[6][61][62] - The optimizer method achieves the highest annualized excess return of 19.0%, but also has the highest volatility and drawdown[70] - Stratified sampling method demonstrates good risk control with an annualized excess return of 7.2% and a maximum drawdown of -8.8%[70] - Regression residuals method has a lower performance with an annualized excess return of 2.7% and a maximum drawdown of -15.9%[70] - The report concludes that while the optimizer method provides the highest returns, it also has higher volatility and less control over tracking error, whereas the stratified sampling method offers better risk control[71][73]