I (a) proved bounds, and (b) showed that those bounds were tight by constructing inputs which came close to the error bounds. The "empirical bounds" demonstrab

I (a) proved bounds, and (b) showed that those bounds were tight by constructing inputs which came close to the error bounds. The "empirical bounds" demonstrably don't work for some inputs.