Indeed. A fastest code competition in a langauge like mIRC's also has an implicit nod towards small, clean code since typically a critical element to obtaining performance is of course to reduce the number of operations and the amount of parsing without making code ambiguous.
I think taking a relatively simple task and asking people to make the fastest code gives people a far greater chance to utilise some skill and is also far more enjoyable and practical. A good example I think was the thread in the Scripting Forum a while back with the $reverse identifier where throughout the thread we went through about half a dozen different implementations and algorithms, getting faster each time until the final solution worked out many times faster than the code that you'd traditionally see used to reverse text.
