Draw a circle, put an equilateral triangle inside it, then pick a chord at random. What is the chance that the chord is longer than a side of the triangle? In 1889, Joseph Bertrand showed that you can get 1/3, 1/2 or 1/4, and every answer looks perfectly valid.
Three fair methods, three different answers
The puzzle appeared in Bertrand’s work Calcul des probabilités. He did not use it to trick students. He used it to show that a popular rule, the principle of indifference, can break down when the possibilities are infinite.
The first method picks two random points on the circle’s edge and joins them. Rotate the triangle so one corner sits on the first point. The chord is long only if the second point lands on the arc facing that corner, which is one third of the circle. Answer: 1/3.
The second method picks a random radius, then a random point along it, and draws the chord at a right angle through that point. The side of the triangle cuts the radius exactly in half, so the chord is long when the point is closer to the center. Answer: 1/2.
The third method picks a random point anywhere inside the circle and uses it as the midpoint of the chord. The chord is long when the point falls inside a smaller circle with half the radius, which has one quarter of the area. Answer: 1/4.
The hidden catch in the word “random”
None of the three calculations contains a mistake. The problem is that “choose a chord at random” does not say how to choose it. The three methods give different weight to different chords, including the ones that pass through the center of the circle.
So the classical conclusion is blunt: without a specified selection method, the question has no unique solution. Other selection methods have even been found that give other numerical answers. In fact, there is an infinite family of them.
A physicist tries to crown a winner
In 1973, Edwin Jaynes published a paper called “The Well-Posed Problem”. His idea was the principle of maximum ignorance: use no information that the problem does not give. Bertrand never specified the size or position of the circle, so Jaynes argued that a fair answer must not care about either.
Imagine throwing straws from far away onto a circle. If you place a smaller circle inside the larger one and slide it around, the pattern of chords on it should not change. Only the second method passes that test. Jaynes used equations to show it was the one and only solution, which gives 1/2.
The experiments that prove everyone right
This is where the story gets odd. Method 2 matches certain physical systems, such as those found in statistical mechanics and gas physics, and it fits the straw-throwing setup. But you can build other experiments that give the other answers.
Attach a spinner to the center of the circle and spin it twice to mark the two endpoints: you get 1/3. Cover the circle with molasses and mark the first spot where a fly lands, then treat it as the chord’s midpoint: you get 1/4. Several observers have run experiments like these and confirmed the results.
Even the fix is up for debate
In 2015, Alon Drory argued that Jaynes’s principle is just as open to interpretation as the principle of indifference. Throwing a dart at the circle and using the hit point as the midpoint produces the third answer, and a spinner-based setup produces the first, each time with the same kind of invariance. The rule meant to settle the debate can be applied in more than one way.
The lesson reaches far beyond circles. Whenever someone says “pick one at random”, the real question is how the picking happens. The same words can hide very different procedures, and the answer you get depends on which one you quietly assume.

