I went to a highly selective university and I certainly could not not nail any algorithm interview with minimal preparation. Algorithms is one class. It consisted mostly of proofs, a little pseudocode, zero working programs. The particular algorithms we spent a lot of time on were pretty obscure in the interview world: Gale-Shapley matching, Ford-Fulkerson max flow. The goal was never to have a wide repertoire of algorithms that you could write down on command, and certainly not anything to do with implementation.