Tool noise
Beginner
Small, random-looking changes in the result when you change a setting a tiny bit.
Novice
The way a design tool’s results jump around when an input changes slightly, for example a small change in the clock target giving a noticeably different area. It makes single comparisons unreliable.
Expert
Place-and-route heuristics are sensitive to small perturbations; results across many near-identical runs look roughly Gaussian. Any ML claim of a few percent needs several seeds or runs per setting to separate the method from the noise.