Bubble Sort vs Insertion Sort vs Selection Sort

Bubble Sort repeatedly swaps adjacent out-of-order pairs, performing many swaps per pass. Insertion Sort shifts elements and inserts each one once, doing far fewer writes. Both reach O(n) on already-sorted data (Bubble needs the swapped-flag optimization), but Insertion Sort's smaller constant factors win on typical input.

For the same number of comparisons, Insertion Sort moves data less and has a tighter loop, so it is roughly 2x faster in practice. This is why hybrid sorts like Tim Sort and Introsort use Insertion Sort — never Bubble Sort — for small subarrays. See the Insertion Sort guide for details.

Insertion Sort and Selection Sort are the closer contest, because they optimise different things.

Insertion Sort wins on almost every axis that normally matters. It is stable, it is adaptive — on nearly-sorted input it approaches O(n) and finishes almost instantly — and it can stop early. Selection Sort is stubbornly indifferent to its input: it performs the same full scan whether the array is random, sorted or reversed, because it must examine every remaining element to be sure it has found the minimum.

So why does Selection Sort survive at all? Because of the one row where it wins decisively: it performs exactly n−1 swaps, the theoretical minimum. Insertion Sort may shift elements O(n²) times. If reads are cheap but writes are expensive or damaging — flash memory with limited erase cycles, or a structure where moving an element is costly — minimising writes can matter more than minimising comparisons. That is a narrow niche, but it is a real one.

The visualiser makes this vivid. Run both on the same array and watch the two counters separately: Selection Sort's comparison count climbs relentlessly while its swap count barely moves, and Insertion Sort does the opposite.

Ranked for general use, it is not close:

• Insertion Sort — the clear winner. Stable, adaptive, simple, and genuinely used in production as the base case inside Tim Sort and introsort once a partition gets small enough.
• Selection Sort — second, and only because of its minimal write count. Otherwise strictly worse.
• Bubble Sort — last. It has no property that another algorithm does not do better. Its enduring value is entirely pedagogical: it is the easiest sorting algorithm to explain and to watch.

An important caveat about all three: quadratic does not mean useless. For arrays of roughly 10 to 20 elements, Insertion Sort typically beats Merge Sort and Quick Sort outright, because the recursion and bookkeeping in an O(n log n) algorithm cost more than the handful of extra comparisons. That is not a curiosity — it is why Tim Sort and introsort both fall back to Insertion Sort for small runs.