Based on ''annual'' averages, a bivariate analysis of the maximum amplitude of the sunspot cycle against its minimum amplitude and the minimum value of the aa geomagnetic index (in the vicinity of sunspot cycle minimum) results in a fit that closely matches the observable record. The bilinear fit has a high coefficient of correlation (r=0.982) and a small standard deviation (s=9.5), suggesting that it may be useful for predicting the size of a sunspot cycle 3 to 4 years before maximum amplitude occurrence. Applying the fit to Cycle 22, one infers an annual average of maximum amplitude Rmax equal to 92¿19 (equivalent to 96¿20, in terms of the 13-month running mean or ''smoothed'' sunspot number).