Bertrand's Paradox — Definition, Formula & Example
Bertrand's Paradox is a probability puzzle in which you choose a "random" chord of a circle and ask how likely it is to be longer than a side of the inscribed equilateral triangle. Three natural methods of choosing the chord give three different answers.
Let a chord be chosen at random in a circle of radius , and let be the event that its length exceeds . The value of depends on the probability distribution placed on the set of chords. Uniform endpoints on the circumference give . A uniform distance from the center along a random radius gives . A uniform midpoint in the disk gives . The problem is ill-posed until the random experiment is specified, because no single distribution is the obvious uniform one.
Key Formula
Where:
- = Length of the chord
- = Radius of the circle
- = Distance from the center of the circle to the chord's midpoint
Worked Example
Problem: A circular dartboard has radius 6 inches. Its inscribed equilateral triangle has side inches. Find the probability that a random chord is longer than the triangle's side, inches, using three different methods of choosing the chord.
Step 1: Turn the length condition into a distance condition. A chord is longer than the triangle's side exactly when its midpoint lies within 3 inches of the center.
Step 2: Method A (random endpoints). Fix one endpoint and pick the other uniformly on the circle. The chord is long only when the second endpoint lands on the arc between the triangle's other two vertices, measured from that first endpoint. This arc spans 120° of the full 360°.
Step 3: Method B (random distance along a radius) picks uniformly from 0 to 6 inches. Method C (random midpoint) picks the midpoint uniformly in the disk, so you compare the area of the inner disk of radius 3 inches with the area of the whole disk.
Answer: The three methods give , , and . Each calculation is correct. They answer different questions, because each one uses a different random experiment.
Why It Matters
Bertrand's Paradox shows why rigorous probability theory requires you to state the sample space and the probability measure explicitly. It also warns against the principle of indifference, which says to treat outcomes as equally likely, when the outcomes form a continuum. This matters in Bayesian statistics when you choose a "noninformative" prior. E. T. Jaynes argued that requiring invariance under rotation, scaling, and translation singles out the answer , but not everyone accepts this as a full resolution.
Common Mistakes
Mistake: Deciding that one of the three answers is the correct one and the others contain errors.
Correction: All three computations are valid. The phrase "choose a chord at random" does not define a probability distribution. You must specify the method before the probability has a single value.
