With a view to find out the most suitable forecasting model for productivity of pearl millet by using past rainfall records of Rajkot district and to identify and quantify the effect of rainfall distribution, 47 consecutive years (1960 to 2006) weekly rainfall data from 23rd to 39th meteorological standard weeks (W), for the cropping season of the pearl millet were collected from Main Dry Farming Research Station, Targhdia (Rajkot). The time trend (T) was also included as an explanatory variable. The district yield of pearl millet (kg/ha) was considered as a response variable. The full model multiple linear regression equation explained 91% variation in pearl millet grain yield. For testing of this fitted model for future, six subsequent years (2001 to 2006) were not included for obtaining prediction equation and verified with actual yields of the district. The deviations between simulated forecasts and actual yields ranged from 2 to 20%. High deviations were due to irregular sowing period, erratic and uneven rainfall distribution of the district. Therefore, high standard error was observed in fitted model. However, this model could be considered for earliest (10 weeks after showing period) forecasting of pearl millet grain yield in Rajkot district.