A junior developer I mentor had to write a Python program for selection sort on a whiteboard, with no sorted() allowed. The answer is a loop that finds the smallest remaining value and swaps it forward, but he kept swapping inside the inner loop.
That bug still sorts, so it slips through. You want one swap per pass.
Quick answer: for each position, find the index of the smallest item from there on, then swap it into place:
for i in range(len(items) - 1):
j = min(range(i, len(items)), key=items.__getitem__)
items[i], items[j] = items[j], items[i]
In this tutorial, I’ll build a Python program for selection sort, trace it pass by pass, sort in descending order and by key, read numbers you type and time it against sorted().
How does selection sort work in Python?
The list has a sorted left part and an unsorted right part. Each pass swaps the minimum of the right part to its front. Every example uses the same six Q1 2026 invoice totals on Python 3.12.5.
def usd(xs):
return ", ".join(f"${x:.2f}" for x in xs)
def selection_sort(items):
n = len(items)
for i in range(n - 1):
smallest = i
for j in range(i + 1, n):
if items[j] < items[smallest]:
smallest = j
items[i], items[smallest] = items[smallest], items[i]
print(f"Pass {i + 1}: {usd(items)}")
return items
invoices = [349.99, 120.50, 89.00, 512.75, 120.50, 64.25]
selection_sort(invoices)
Output:
Pass 1: $64.25, $120.50, $89.00, $512.75, $120.50, $349.99
Pass 2: $64.25, $89.00, $120.50, $512.75, $120.50, $349.99
Pass 3: $64.25, $89.00, $120.50, $512.75, $120.50, $349.99
Pass 4: $64.25, $89.00, $120.50, $120.50, $512.75, $349.99
Pass 5: $64.25, $89.00, $120.50, $120.50, $349.99, $512.75

The outer loop runs n - 1 times; the last item’s already in place. The inner loop only tracks an index, and the swap is plain tuple unpacking, like when you swap two numbers in Python. Pass 3 swapped $120.50 with itself.
Selection sort in descending order
To put the biggest invoice first, track the largest value and flip the comparison:
if items[j] > items[largest]: # was <
largest = j
On the same invoices that printed $512.75 first and $64.25 last. If you’ve already sorted ascending, you can reverse the list in Python instead.
Is selection sort stable? Sorting records by a key
For rows of data, I gave the function a key argument, like sorted() has, and sorted four orders:
def selection_sort(items, key=lambda x: x):
n = len(items)
for i in range(n - 1):
best = i
for j in range(i + 1, n):
if key(items[j]) < key(items[best]):
best = j
items[i], items[best] = items[best], items[i]
return items
orders = [
("Olivia Carter", "TX", 120.50),
("Noah Bennett", "CA", 120.50),
("Ethan Brooks", "OH", 89.00),
("Mia Ramirez", "TX", 64.25),
]
by_total = lambda r: r[2]
results = {"selection_sort:": selection_sort(orders[:], by_total),
"sorted():": sorted(orders, key=by_total)}
for label, rows in results.items():
print(label)
for name, state, total in rows:
print(f" {name:<14} {state} ${total:.2f}")
Output:
selection_sort:
Mia Ramirez TX $64.25
Ethan Brooks OH $89.00
Noah Bennett CA $120.50
Olivia Carter TX $120.50
sorted():
Mia Ramirez TX $64.25
Ethan Brooks OH $89.00
Olivia Carter TX $120.50
Noah Bennett CA $120.50

Selection sort isn’t stable: the long swap moved Olivia Carter behind Noah Bennett. sorted() keeps ties in order, which you’ll need when you sort a list of tuples by two fields.
Selection sort on numbers the user enters
This version reads prices with input() and converts each one with float():
# usd() from the first example
# selection_sort() from the first example, minus the print
raw = input("Enter prices separated by spaces: ")
prices = [float(p) for p in raw.split()]
print("Sorted:", usd(selection_sort(prices)))
Output:
Enter prices separated by spaces: 19.99 5.49 42.00 7.25 12.80
Sorted: $5.49, $7.25, $12.80, $19.99, $42.00

How fast is selection sort compared with sorted()?
It always makes n(n-1)/2 comparisons, so it’s O(n²) even on sorted input, with at most n – 1 swaps. Here’s 2,000 random prices:
import random
import timeit
# selection_sort() from the first example, minus the print
random.seed(2026)
data = [round(random.uniform(5, 900), 2) for _ in range(2000)]
t_sel = timeit.timeit(lambda: selection_sort(data[:]), number=5) / 5
t_py = timeit.timeit(lambda: sorted(data), number=5) / 5
print(f"selection_sort: {t_sel * 1000:.1f} ms")
print(f"sorted(): {t_py * 1000:.2f} ms")
print(f"sorted() was {t_sel / t_py:.0f}x faster")
Output:
selection_sort: 72.9 ms
sorted(): 0.23 ms
sorted() was 321x faster

sorted() runs Timsort in C. On nearly sorted data, insertion sort is the better simple sort.
Common selection sort mistakes and how to fix them
Swapping inside the inner loop. That’s a slower exchange sort. Move the swap after the scan.
Expecting a new list. It sorts in place, so pass items[:] if you’ll need the original.
To compare it with other hand-written sorts, start with these:
- Sorting algorithms in Python
- Python program for bubble sort
- Sort a list without the sort() function
- Sort a list of objects by attribute
Frequently asked questions
What is the time complexity of selection sort?
O(n²) in every case, because it always scans the whole unsorted part. Extra space is O(1).
Is selection sort stable?
No. A swap can jump an item past an equal one. Use sorted() when tie order matters.
How many swaps does selection sort make?
At most n – 1, one per pass.
What’s the difference between selection sort and bubble sort?
Bubble sort swaps neighbors many times per pass; selection sort swaps once.
Can selection sort sort strings?
Yes, alphabetically, since Python compares strings with <. Mixing strings and numbers raises TypeError.
Where can I read about sorting in Python?
The official Sorting Techniques HOWTO covers keys and stability.

Bijay Kumar is a 13-time Microsoft MVP with more than 18 years in software development, and the founder of Python Guides and TSinfo Technologies. He started out building .NET and SharePoint solutions at HP, TCS and KPIT before moving into Python, machine learning and AI, and he also builds web apps with TypeScript and React. He writes the tutorials here himself, and every example is run before publishing so you see the real output. More about Bijay · Microsoft MVP profile · LinkedIn