In fact I don't think his conclusion is sound. If you're into algorithmic competitions you most likely don't need to change the language. If you devise a solution with a good enough complexity, the competition (if well designed) should allow enough time for the execution of that (near-)optiumum solution, while not awarding points for solutions which are asymptotically worse. The time taken by algorithms with different complexities vary with the size of the input, as a nonconstant function, while implementing the same algorithm in different languages will give you a difference of a constant factor.

What I'm trying to say, (stating the obvious): When the input is large enough, it doesn't matter what language you're programming in. And programming competitions should (and generally do) only focus on that.

But I'll admit that I'm a huge fan of optimizing my algorithms with bit-level operators, extreme care of memory allocation and other tools C offers.