Why time only runs forwards

A film of two billiard balls colliding looks perfectly normal played backwards. So does a planet orbiting, a pendulum swinging, two molecules bouncing off each other. The microscopic laws of physics don't care which way time runs; reverse every velocity and the motion is still a legal solution. And yet. Eggs don't unscramble. Coffee doesn't un-stir. A smell doesn't gather itself back into the bottle. The big picture has an unmistakable direction even though every tiny piece of it is reversible.

That's a real puzzle, and it bothered the people who first thought about it properly. If nothing in the rules picks a direction, where does the arrow of time come from? The answer, which I find one of the most satisfying ideas in physics, is that it isn't in the rules at all. It's in the counting.

Take the partition out

Here's the whole thing in a box. Teal particles on the left, warm ones on the right, a partition between them. Pull the partition and let the particles do what particles do: drift, bounce off the walls, mind their own business. Watch what the box does as a whole.

It mixes, every time, and the entropy readout climbs and then sits at the top. Now here's the uncomfortable question: which rule did that? None of them. Each particle is just moving in a straight line and bouncing elastically, all of it perfectly reversible. No particle "wants" to spread out. The box mixes anyway. So the direction has to come from somewhere other than the per-particle physics.

It's a counting argument

The trick is to separate two things we usually blur together. A microstate is the complete description: every particle's exact position and velocity. A macrostate is the blurry view: roughly how many teal are on the left, how mixed it looks. The point is that one macrostate corresponds to a colossal number of microstates, and crucially, not the same number for every macrostate.

Ask how many ways you can arrange the particles so they look "mixed" versus "all teal on the left". The separated arrangement is rare and special: there's basically one way to have every teal particle on the left. The mixed arrangement is generic: there are astronomically many ways to scatter the colours evenly. Entropy is just the log of that count:

$$S = k_B \ln W$$

Copy LaTeXwith $$ delimitersCopy mathexpression only

For our two-colour box, the number of ways to split $N$ particles so that $n$ are on the left is a binomial coefficient, and the count is overwhelmingly peaked at the even split:

$$W(n) = \binom{N}{n} = \frac{N!}{n!\,(N-n)!}$$

Copy LaTeXwith $$ delimitersCopy mathexpression only

So the box doesn't mix because of a law that says "mix". It mixes because there are so many more mixed-looking microstates than separated ones that, wandering randomly through its options, it spends essentially all its time in the mixed ones. The second law of thermodynamics, the one that says entropy never decreases, isn't a force pushing things around. It's a statement about counting that's true with overwhelming probability.

The reversibility paradox

If you're feeling sharp you might object: hang on, the rules are reversible, so the un-mixing motion exists. Reverse every particle's velocity at the mixed moment and the box would obediently un-mix, particle by particle, back into neat halves. You're completely right, and this is exactly the objection Loschmidt raised to Boltzmann in the 1870s. It's not a flaw in the argument; it's the heart of it.

But where did the order come from?

There's one more turn of the screw, and it's the part that still genuinely puzzles me. The counting argument explains why entropy goes up from any given moment. But by the same logic it should have been higher in the past too: the maths is symmetric, so "more microstates" wins looking backwards as much as forwards. Run the reasoning naively and it predicts the past was also high-entropy, which is plainly false: yesterday was more ordered than today, and the early universe was extraordinarily ordered.

So, why does time run one way?

Because "ordered" is rare and "messy" is common, and a system left alone wanders from the rare into the common simply because there's so much more of it. Stir the counting argument into a universe that happened to start tidy, and you get every irreversible thing you've ever seen: heat flowing hot-to-cold, smells spreading, the unscrambleable egg, and the plain fact that you can remember breakfast but not dinner. The micro-laws never picked a direction. We did, by living downstream of a low-entropy beginning, in a world where the messy outcomes outnumber the tidy ones by numbers with twenty-three zeros.

Reading further

• Boltzmann (1877): where $S = k \ln W$ comes from. The equation is literally carved on his gravestone. Any statistical-mechanics text traces the original argument.
• Feynman, The Character of Physical Law, chapter 5 ("The Distinction of Past and Future"): the clearest hour you can spend on exactly this question, from someone who thought about it beautifully.
• Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity: the modern textbook treatment, free online, with the counting made precise. sethna.lassp.cornell.edu
• Carroll, From Eternity to Here: a whole popular book on the arrow of time and the past hypothesis, if the last callout left you wanting more.