Abstract: The statistical downscaling technique is used to predict rainfall using general circulation model (GCM) data. One important factor to improve prediction accuracy is selection of GCM grid resolution. The purpose of this study is to determine the best GCM grid resolution to predict rainfall. The data used was GCM data from CFSRv2 with grid resolutions of 0.312°"×" 0.312°; 0.5°"×" 0.5°; "2.5°×2.5°" and local rainfall data in West Java. The determination of best grid resolution began with determining the GCM domain in each rainfall observation station based on minimum correlation value of 0.3 between GCM data and local rainfall data. The domains were evaluated by LASSO regression based on the smallest RMSEP and the largest correlation. The results showed that the best grid resolution is "2.5°×2.5°" .