Clustering on a ring

Drop 4 points randomly on a circle. What are the chances they all land in the same half?

Your first instinct might be: the circle is symmetric, so place one point anywhere (since any point, probability is 1) and ask whether the other three land in the same semicircle. Each of the 3 remaining points independently has a $1/2$ chance of landing in that half, so $P = (1/2)^3 = 1/8$.

Let's try it:

But that's wrong. The semicircle isn't fixed in advance. It can be anchored at any of the four points. Drop 4 points and click each one to see which anchor works:

Place points and check:

Try different values of $N$. For $N = 2$, the answer is 100% (any two points fit in a semicircle). For $N = 3$, it's about 75%. As $N$ grows, the probability drops sharply.

The counting argument

Label the points $1, 2, \ldots, N$ clockwise around the circle. For each point $i$, define the event:

$E_i = \text{"all other points lie in the clockwise semicircle starting at point } i\text{"}$

Each $E_i$ has probability $(1/2)^{N-1}$: once we fix point $i$ as the anchor, each of the remaining $N-1$ points independently has a $1/2$ chance of landing in that semicircle. There are $N$ such events. If we could just add them up, we'd get $N / 2^{N-1}$.

But can we? Adding probabilities only works when events are mutually exclusive (if one event happens, it automatically excludes the other from happening). The key claim is:

Why the events can't overlap

Suppose event $E_i$ occurs: all points lie in the clockwise semicircle starting at point $i$, meaning the arc from $i$ clockwise to the point just before $i$ covers at most 180 degrees. Place some points and click one to see its semicircle and the gap:

Notice the gap — when one anchor works, the counterclockwise arc back to it is at least 180°. That gap is exactly what prevents any other point from being an anchor. Step through the proof below to see why:

Here's the same argument written out formally:

Since the events are mutually exclusive:

$P(\text{all in some semicircle}) = P\Large\left(\bigcup_{i=1}^{N} E_i\right) = \Large\sum_{i=1}^{N} P(E_i) = N \times \Large\frac{1}{2^{N-1}} = \Large\frac{N}{2^{N-1}}$

Extending to smaller arcs

The same argument works for any arc, not just semicircles. If the arc has length $x$ times the circumference (where $x \leq 1/2$, and for semicircles $x = 1/2$):

$P(\text{all } N \text{ points in some arc of length } x) = N \cdot x^{N-1}$

Adjust the arc size and number of points to see how the probability changes:

Competition appearances

This problem shows up frequently:

• Putnam style: "N points chosen uniformly on a circle. Find the probability they lie within an arc of angle $\theta$."
• IMO training: The mutual exclusivity argument is a model for how geometric symmetry simplifies counting
• Quant interviews: Jane Street and Two Sigma use this as a warm-up to test geometric probability intuition
• Generalization: The same technique applies to points on a sphere (what's the probability $N$ points fit in a hemisphere? Answer: $N / 2^{N-1}$ in 2D; the 3D case is harder)

The takeaway

This problem demonstrates the technique of decomposing a complex event into simpler pieces and checking whether those pieces overlap. When they don't (mutual exclusivity), you just add. The geometric insight that "only one point can anchor a covering semicircle" is what makes the addition valid. In competition math, whenever you can decompose a probability into $N$ identical, non-overlapping cases, the formula $N \times p$ gives you the answer instantly.