Perceptron learning is discussed in the context of socalled scoring systems used for assessing creditworthiness as stipulated in the Basel II central banks capital accord of the G10-states. The solution of a related ranking problem using a generalized version of the pocket algorithm is described. A correctness proof of the algorithm is given. It is argued that the reuslts obtained may be exploited to compute associated probabilities using a logistic activation function and maximum likelihood methods. Some preliminary experimental results are exhibited.