Though the worst case performance of an algorithm may be f(n), the average worst case over n runs may be asymptotically less than n*f(n).

We will study three methods for average worst case analysis:

- 1.
**Aggregate Method.**Show all n runs take a total worst case time T, thus the average worst case is T/n.- 2.
**Accounting Method.**Performance costs assigned to some of the n runs are overestimates and are used as credit for underestimates of other runs.- 3.
**Potential Method.**The overestimates add to the ``potential energy'' of the method.