Abstract Background Prior study has demonstrated that the biplane single-beat method could be used to assess left ventricular function during atrial fibrillation at a beat with equal subsequent cycles. The study was to test whether we could improve the method by measuring a few beats with equal subsequent cycles and cycle-length limits. Methods In 75 patients with atrial fibrillation, stroke volume and ejection fraction were determined from simultaneous biplane views of left ventricle for 20 beats using a matrix-array transducer and a biplane Simpson's rule. The influence of cycle lengths on the values of systolic parameters at beats with equal subsequent cycles was examined from the plot of normalized parameters (measured values / average values) against cycle lengths. The values of 1 to 3 beats with equal subsequent cycles and cycle-length limits were averaged and compared with the average values over 20 beats by Bland–Altman and mean percentage difference analysis. The variability of repeat measurements was evaluated in 10 patients. Results The systolic parameters measured at beats with cycle lengths shorter than 500 ms were usually far below the average values. Agreement and mean percentage difference analysis revealed improved accuracy when 2 or 3 beats with cycle-length limits (> 500 ms) were used for assessment. As the variability of averaging 2 or 3 beats is no greater than that of repeat measurements, both methods are equally good. Conclusions Accurate assessment of left ventricular systolic function in atrial fibrillation can be obtained by averaging 2 beats with equal subsequent cycles and cycle-length limits (> 500 ms).