Within a group of students of similar scores, a “normal” distribution should be a narrowing range of probabilities. However, because the probabilities is exponential (to the number of identical answers), it is misleading to just look at the probabilities, as their range will vary significantly.
It is better to look at the logarithmic values. If the logarithmic values seem to be somewhat clustered (after sorting them), then things are probably okay. However, if some logarithmic value is out of the cluster, then those cases should be examined with more care.