Understanding Bertrand's Paradox
Bertrand's paradox is a deceptive problem in probability theory. It begins with intuitive geometry: place an equilateral triangle inside a circle such that all three vertices touch the circumference. The triangle's sides are chords of the circle. The paradox asks: if you draw a chord at random, what is the probability that it will be longer than a side of the inscribed triangle?
What makes this a paradox is not the question itself, but the answer. There is no single answer. Depending on your method of selecting the random chord, you can arrive at three different probabilities—all mathematically defensible. This discovery troubled mathematicians in the 19th century because it suggested that even well-posed mathematical problems could yield multiple correct solutions.
The root cause is the phrase "draw a chord at random." Randomness over an infinite space requires explicit definition. Without specifying how the randomness is distributed—that is, which sampling procedure you employ—the problem remains ambiguous. This is where the principle of indifference, which assigns equal likelihood to all outcomes, becomes insufficient.
The Three Solutions
Bertrand's paradox yields three distinct probabilities depending on the method used to generate random chords:
Random Endpoints Method: P = 1/3 ≈ 0.3333 or 33.33%
Random Radial Point Method: P = 1/2 = 0.5000 or 50%
Random Midpoint Method: P = 1/4 = 0.2500 or 25%
P— The probability that a randomly drawn chord is longer than the inscribed triangle's sideMethod— The procedure used to select the random chord: endpoints, radial points, or midpoints
The Three Sampling Methods Explained
Random Endpoints Method: Select one vertex of the inscribed triangle as a fixed starting point. Draw a chord from this point to a random location on the circle's circumference. A chord is longer than the triangle's side only if its endpoint falls within a specific arc—exactly 1/3 of the circumference opposite the starting vertex. Hence, the probability is 1/3 or 33.33%.
Random Radial Point Method: Choose a random point along a radius of the circle (from centre to edge). Construct a chord perpendicular to this radius through that point. The key insight: a chord is longer than the triangle's side if and only if its midpoint lies within the inscribed circle. Since the inscribed circle occupies exactly half the area of the original circle, the probability is 1/2 or 50%.
Random Midpoint Method: Pick a random point anywhere inside the original circle. This point becomes the midpoint of your chord (the chord perpendicular to the radius through that point). A chord exceeds the triangle's side length when its midpoint falls inside the inscribed circle. The inscribed circle's area is 1/4 that of the original circle, yielding a probability of 1/4 or 25%.
Key Pitfalls and Insights
Understanding Bertrand's paradox requires careful attention to these practical considerations:
- Sampling Ambiguity Is Real — The paradox is not a mathematical error but a genuine reminder that "random" over infinite spaces is meaningless without specification. In practical applications—simulation, experimental design, Monte Carlo methods—always make your sampling procedure explicit.
- The Principle of Indifference Has Limits — Treating all outcomes as equally likely seems fair, but over infinite domains it leads to contradictions. This principle works well for finite discrete spaces but breaks down when applied naively to continuous distributions.
- Symmetry Does Not Resolve It — Even though the circle and inscribed triangle are geometrically symmetric, this symmetry alone cannot pick a unique answer. The three methods are equally symmetric in their own geometric frameworks, yet they produce different results.
- Applications Beyond Theory — Bertrand's paradox appears in physics (e.g., radiation problems), engineering (tolerance analysis), and statistics (Bayesian priors). Recognizing the ambiguity in problem statements can prevent misinterpretation of results in real-world analyses.
Why Multiple Answers Are All Valid
The crucial insight is that each of the three solutions emerges from a different, internally consistent probability distribution over the infinite space of chords. The endpoints method, the radial method, and the midpoint method each impose a different notion of uniformity.
This raises a philosophical question: which distribution is "correct"? The answer depends on the physical or practical context. If you are generating chords by dropping sticks onto a circle, the physics of the process determines which solution applies. If you are a mathematician working purely abstractly, you must state your assumption explicitly.
The paradox thus serves as a cautionary tale: infinite spaces demand rigorous specification of probability measures. Without it, ambiguity is not just possible—it is inevitable. Modern probability theory addresses this through measure theory, which formalizes exactly how to assign probabilities to events in infinite spaces in a consistent manner.