Photometric optical data of QSO 0957+561 covering the period 1984-1999 are analyzed to discern between the two values of the time delay (417 and 424 days) mostly accepted in the recent literature. The observations, performed by groups from three different institutions-Princeton University, Harvard-Smithsonian Center for Astrophysics, and Instituto de Astrofísica de Canarias-and including new unpublished 1998-1999 data from the IAC80 Telescope, were obtained in five filters (V, R, I, g, and r). The different light curves have been divided into observational seasons, and two restrictions have been applied to calculate the time delay better: (1) points with a strange photometric behavior have been removed, and (2) data sets without large gaps have been selected. Simulated data were generated to test several numerical methods intended to compute the time delay (ΔτAB). The methods giving the best results-the discrete correlation function, δ2, z-transformed discrete correlation function, and linear interpolation-were then applied to real data. A first analysis of the 23 different time delays derived from each technique shows that ΔτAB must be in the interval 420-424 days. From our statistical study, a most probable value of ΔτAB=422.6+/-0.6 days is inferred.