The goal of the present work is to reduce the number of thetraining samples used in our color prediction model based onCIEXYZ using an Effective Coverage Map while keepingsatisfying prediction. A general approach is proposed in thispaper to choose the best reference combination for the trainingsamples. The approach is based on the dot gain behavior ofeach primary ink, which is characterized by three curves usingCIEXYZ tri-stimulus values. The proposed approach is built inour model to predict the color values for the color prints usingtwo different devices, i.e. a laser printer and an inkjet printer.For the laser printer the number of the training samples isreduced from 125 to 64 while still giving quite good result. Theapproach also shows that for the test laser printer it is possibleto further cut this number to 53 with a satisfying result. For theinkjet printer the number of training samples for our model isreduced from 125 to 79 or 64, both giving satisfying results.