negative scoring values for training would currently be trail and error for users. Training should test a range of negative values automatically to find optimal