This study compiles data on actively managed mutual funds in China spanning from 2007 to 2020 to construct portfolios of superior funds using the Fund Confidence Set (FCS) algorithm. Time-varying alphas and betas are estimated using an adaptive unscented Kalman filter with maximum posterior and random weighting rather than the extended Kalman filter. Moreover, an exponential smoother is integrated into the FCS algorithm, leading to improvements in portfolios’ performance. These two modifications to the FCS algorithm further enhance the performance of superior fund portfolios, producing an average monthly excess return of up to 1.88 %.