How to Choose Fairly Between Three Options With a Coin or Die
A fair three-way choice assigns probability 1/3, or 33.33%, to each option before any random result is generated.
A worked example with numbers
Suppose three roommates must choose Film, Walk, or Board Game. A six-sided die supplies six equally likely faces. Map faces 1–2 to Film, 3–4 to Walk, and 5–6 to Board Game. Each activity owns two of six faces, so each has 2/6 = 1/3. The mapping must be declared before rolling. If the die shows 4, Walk wins because face 4 belongs to the 3–4 group, not because somebody liked the result after it appeared.
How to set the rule before the result
A coin can also make three equal choices, but it needs a rejection rule. Flip twice: HH selects Film, HT selects Walk, TH selects Board Game, and TT means “discard and flip two coins again.” The first three two-flip patterns each have probability 1/4. Conditional on not getting TT, each accepted pattern has (1/4) ÷ (3/4) = 1/3. The rejected TT outcome is not a secret redraw; it is part of the announced mechanism. A single die roll is usually easier for a group to audit, while a number wheel can show the three labels directly.
Common mistakes that change the odds or the process
Do not map heads to option A and tails to option B, then “flip again for C if necessary.” That creates an undefined rule because there is no necessary coin outcome. Do not give one option faces 1–3 and the others faces 4–5 and 6: the probabilities are 1/2, 1/3, and 1/6. Do not use `roll % 3` on a die numbered 1 through 10 unless the implementation rejects unmatched values; ten is not divisible by three, so raw remainders make one class appear four times and the others three times.
Where this method stops being appropriate
The method is useful only when the options are genuinely interchangeable after constraints have been checked. A three-way restaurant choice is fine after allergies, budgets, and opening hours are settled. It is not a substitute for choosing emergency action, consent, money allocation, or a decision where one option could harm someone. Randomness allocates a tie; it does not evaluate consequences.
How the random source fits into the rule
For software, the important numerical condition is that the source range can be partitioned into equal-size groups. MDN notes that MDN’s Math.random reference returns a pseudorandom floating-point number in the interval from 0 inclusive to 1 exclusive; it also warns that it is not a cryptographic source. Browser code that needs stronger unpredictability can obtain integer values from MDN’s Crypto.getRandomValues reference.
Choosing the mechanism before labels acquire meaning
Three options are a useful test because a coin cannot represent one third in a single flip. A six-face die works because six is three times two; the pairs 1–2, 3–4, and 5–6 are equally large. A nine-number source also works with three groups of three. A five-face source cannot be divided into three equal integer groups, so it needs a stated discard or a different source. Write the mapping beside the labels before anyone rolls. Changing “Film is 1–2” after a 2 appears is exactly the kind of invisible discretion the mapping removes.
The two-coin method has a measurable cost. TT appears with probability 1/4, so accepted results arrive after 1 ÷ (3/4) = 4/3 two-flip attempts on average. That is still fair when TT is announced as a discard, but a die is usually simpler in a room. If the decision needs a ranked first, second, and third choice, shuffle the three labels instead; one winner and one complete ordering are different sample spaces.