Pick up a square tile. Close your eyes. Ask a friend to either leave it alone, or spin it exactly 90, 180, or 270 degrees, or flip it over along one of its lines of symmetry, then open your eyes.
Can you tell what they did?
For a plain square, you can't. Every one of those moves lands it back looking exactly the way it started. That small, almost silly observation is the seed of one of the most useful ideas in all of math: the group.
Mathematicians took that observation and asked a sharper question. What if we forget the square itself and study the list of moves? It turns out that list has its own structure, rules that show up again and again: in clock arithmetic, in Rubik's cubes, in crystals, in the equations of particle physics. That structure is a group.
A symmetry is a move you can't detect
Let's make "you can't detect it" precise. A symmetry of a shape is any move, rotation, reflection, whatever, that lands the shape back on top of its own outline. Not just close. Exactly. Every point of the shape has to end up covering some point that used to be part of the shape.
The square's outline never moves. Track which lettered corner ends up where.
Click through the buttons above. Watch which lettered corner lands where. A square has exactly 8 symmetries: doing nothing, three rotations, and four reflections. Not 7, not 9. Try to imagine a rotation by 45 degrees instead. It would tilt the square into a diamond, sticking out past its own original outline. That move doesn't count, because the square only has four-fold rotational symmetry.
Notice something else. Every button keeps the square's outline looking untouched, but the letters (which a real square's identical corners would hide) tell you something actually happened underneath. That hidden bookkeeping, tracking what really moved even though the outline looks unchanged, is the whole game in group theory.
You can never combine your way out of the list
Here's a question worth sitting with. If you rotate the square 90 degrees and then reflect it, is the combined move also one of the 8 symmetries on our list, or something new entirely?
Try it by hand first. Rotate a square piece of paper, then flip it. It's still sitting exactly on its own outline. So whatever that combined move was, it has to already be one of the 8 things we found. There's no hidden 9th one waiting to appear.
Click a cell: row move, then column move. Every row and column contains all six symmetries exactly once.
This grid is called a Cayley table. Every cell is what you get from combining the row's move with the column's move, in that order. Click around. You will never land on an empty cell, and you will never land on a symmetry outside our familiar list. Mathematicians call this property closure: combine two elements of a group, and the result is guaranteed to still be in the group.
Look even closer. Every row and every column contains each symmetry exactly once, never twice, never zero times. That's not a coincidence, it's forced. And it has a lovely side effect: since "do nothing" shows up exactly once in every row, every single move has some other move that undoes it completely. Every symmetry has an inverse, guaranteed, just from that one pattern.
Rotating first isn't the same as flipping first
Quick question. If you rotate a triangle and then flip it, do you land in the same place as flipping it first and then rotating?
Socks, then shoes, is not the same as shoes, then socks. Order can matter, and it's worth checking whether it matters here too.
Pick two moves. Watch whether doing them in one order lands the same as the other.
Pick two moves above and watch both orders play out side by side. Pick two rotations, r and r², and the order never matters, both paths land in the same spot. Now pick a rotation and a reflection instead, and the two paths split apart, landing at different corners entirely.
Groups where order never matters are called abelian. Groups where it sometimes matters, like the symmetries of a triangle or a square, are called non-abelian. Both kinds are equally valid groups. Nothing in the definition of a group requires order to be irrelevant, that's an extra bonus property some groups happen to have.
Numbers can be symmetries too
Everything so far has been shapes. But the exact same structure shows up in places that have nothing to do with rotating cardboard.
Think about a clock. 10 o'clock plus 4 hours is 2 o'clock, not 14 o'clock. The hours wrap around. That wrapping is called modular arithmetic, and it behaves exactly like our square's symmetries: combine two hours and you get a valid hour back (closure), there's a "do nothing" hour (adding 0), and every hour has one that undoes it (adding its complement gets you back to 12, or 0, depending how you count).
Set the clock size and step, hit play, and watch it wrap back around to 0.
Set the clock size and the step size, then hit play. Watch the marker walk around the dial, wrapping past the top back to 0 whenever it overflows. For some step sizes it visits every number before returning home. For others it only visits a handful before landing back on 0 and repeating the same short loop forever.
That number of steps before returning to 0 is called the order of the element, and it's controlled by a much older idea: the greatest common divisor of the step size and the clock size. Set the clock to 12 and the step to 5, and you'll visit all 12 hours before landing home. Set the step to 4 instead, and you'll only ever visit 3 of them, because .
Four rules are all it takes
We've now seen two groups that look nothing alike on the surface, a square's symmetries and a clock's hours, sharing the exact same underlying skeleton. Time to write that skeleton down properly.
A group is a set together with a way of combining two elements, written , satisfying four rules:
Every one of those rules already showed up, in pictures, before we ever wrote a symbol. Closure was the Cayley table never having an empty cell. Identity was "do nothing." Inverses were every row containing the identity exactly once. Associativity, the fact that grouping parentheses doesn't matter, is quietly true for both symmetries and clock arithmetic too. We never needed to point at it directly, because sequences of physical moves and sequences of additions never care how you bracket them.
Notice what's missing from that list. Commutativity isn't one of the four rules. is optional, a bonus property some groups have, like our clock, and others don't, like our square.
Groups are hiding everywhere
Once you know what to look for, groups show up constantly. The scrambled moves of a Rubik's cube form a group, with "do nothing" as the solved state and every scramble having some sequence that undoes it. Crystals form groups based on which rotations and reflections leave their lattice looking unchanged, which is how physicists classify possible crystal structures. Cryptography leans on groups built from clock-style arithmetic over enormous numbers, where combining two elements is easy but undoing a specific combination is brutally hard.
None of these examples were designed to resemble each other. They just all obey the same four rules, closure, associativity, identity, inverses, and that's enough for one theory to describe all of them at once.
The short version
A group is a set of moves, or numbers, or actions, that you can combine and always land back inside the same set (closure). There's always a do-nothing move (identity), every move can be undone (inverses), and chaining moves together doesn't care how you bracket the chain (associativity).
Order sometimes matters, non-abelian, like a square's rotations and flips, and sometimes doesn't, abelian, like a clock's hours. Both are still groups either way.
That's it. Four rules, and from a plain square tile and a wall clock, an entire branch of math falls out, one that ends up describing crystals, cryptography, and the deep symmetries of physics itself.