Let’s say there are 100 records in the training dataset. The observations are arranged in decreasing order of probability ...