This is more or less my experience in bioinformatics, computer vision, and machine learning. MATLAB is much better than people without MATLAB experience believe. It’s also much worse than people with _only_ MATLAB experience believe. From a robust package perspective, MATLAB is probably more complete than non-statistical R, but in recent years it’s lagged far behind the scientific Python ecosystem. It’s difficult to imagine it catching up to an alternative that’s free and better (from a completeness, performance, or user experience perspective).