LZ4 is amazing - it's basically the same idea as LZ77, only implemented carefully, with all the dozens parameters and tradeoffs made in a certain way. Couple this with modern amounts of brute force thrown in an optimal parser (LZ4HC) and you get something quite extraordinary, on the "Pareto frontier" of compression efficiency. There's a tendency to focus on algorithms and scoff at implementation details, but LZ4 was possible as an algorithm 30 years ago, yet the implementation details crystallized to perfection only, say, five years ago.