The quadratic sort with the most predictable cost — and no early exit.
Selection sort scans the unsorted region for its minimum and swaps that minimum into the first unsorted position. It then never looks at that position again and moves on to the next one.
Unlike bubble and insertion sort, the amount of work barely depends on the input. Even a sorted array costs the same as a reversed one, because the minimum still has to be found by scanning the whole remaining region every pass.
The cost profile is the interesting part. Selection sort performs about n squared over two comparisons — the most of any sort here — but at most n minus one data moves, since each pass swaps at most once.
That makes it attractive when moving an element is dramatically more expensive than comparing two, such as on flash storage or over a network. In those settings the number of writes dominates, and selection sort's write count is the best available.
Selection sort is not stable. A long-distance swap can move an equal element past an element that should come after it, so the relative order of equal keys is not preserved.
The visualiser uses distinct values, so instability is invisible there. It is a property of the algorithm rather than of the animation, and it is worth remembering when choosing a sort for records that carry a secondary key.