Monty Hall Problem: Why Switching Wins Two Times Out of Three

A stage illustrating choices in the Monty Hall problem

Short answer: in the standard Monty Hall problem, switching is the better strategy. Staying with your first door wins with probability 1/3; switching wins with probability 2/3. That conclusion depends on a host who knows where the prize is, always reveals a goat, and always offers the switch.

The puzzle feels like a fifty-fifty choice because only two closed doors remain. Yet those doors did not survive the same process. You selected one without information. The host deliberately protected the other while removing a losing option. Probability follows that information-generating process, not the number of objects left on stage.

The Monty Hall problem in one round

Three doors conceal one car and two goats. You choose Door 1. Before revealing your choice, the host opens one of the other doors and shows a goat. He then offers a choice: keep Door 1 or move to the only other closed door.

Steve Selvin presented the mathematical puzzle in letters to The American Statistician in 1975. The familiar television framing makes the question memorable, but the result comes from a precise model. If the host acts under a different policy, the answer may change.

Why switching wins two times out of three

Your first choice has a 1/3 chance of containing the car. Therefore, the pair of doors you did not choose carries the remaining 2/3 chance. The informed host removes a goat from that pair. Because he cannot reveal the car, the pair’s 2/3 probability effectively concentrates on the one unopened door.

Initial choiceProbabilityStaySwitch
Car1/3WinLose
Goat2/3LoseWin

This gives the cleanest rule: switching wins whenever the initial choice is wrong. A random first choice is wrong in two of the three equally likely prize locations. Stanford’s conditional-probability lecture writes the same result as P(win by switching) = 1/3 × 0 + 1/3 × 1 + 1/3 × 1 = 2/3.

Enumerating all three prize locations

Suppose you always pick Door 1. If the car is behind Door 1, switching loses. With the car behind Door 2, the host must reveal Door 3 and switching wins. When the car is behind Door 3, the host must reveal Door 2 and switching also wins. There are two wins among three equally likely cases.

Why the last two doors are not fifty-fifty

Counting two closed doors ignores selection. The host’s action is constrained: he knows the prize location and must reveal a goat. The remaining unchosen door survived a filter that your first door never faced.

Consider a different experiment. A person who does not know the prize location opens one unchosen door at random and happens to reveal a goat. That event could have gone wrong by exposing the car. Conditional probabilities now depend on the random opener’s policy. The sentence “a goat was revealed” does not fully describe the evidence; the method matters.

Assumptions behind the standard answer

  • The car is placed randomly, giving each door prior probability 1/3.
  • The host knows which door contains the car.
  • He always opens an unchosen goat door.
  • He always offers the contestant a switch.
  • When two goat doors are available, his tie-breaking rule does not encode an undisclosed signal.

These assumptions are not technical decoration. Imagine a host who offers a switch only when your first choice is correct, or one who occasionally opens doors at random. The observed offer would then carry different information. A careful solver states the host policy before quoting 2/3.

A theatre curtain illustrating hidden information in the Monty Hall problem
The host’s knowledge—not merely an opened curtain—changes how the information should be interpreted. Photo: DJ Paine/Unsplash.

A classroom simulation that makes the result visible

Make three cards: one car and two goats. One student shuffles and places them face down. A second student picks a card. The first student, who can see all three, reveals an unchosen goat and offers the switch. Repeat the game while the player always switches.

Thirty rounds will rarely produce exactly twenty wins. Small samples wobble. With hundreds or thousands of rounds, however, the switching win rate tends to move toward 2/3. Simulation supplies evidence and intuition; enumeration supplies the proof.

A short program follows the same structure. Randomize the car door, randomize the contestant’s first choice, let the host select a legal goat door, switch to the last closed door, and record the outcome. Run that loop many times. Compare it with a second player who always stays.

The 100-door version

Scale the game to 100 doors. You select one, giving your choice only a 1/100 chance of being correct. The host opens 98 goat doors and leaves your door plus one other door closed. Staying keeps the original 1/100 chance. Switching captures the 99/100 chance that your first choice was wrong.

The large version does not introduce a new principle. It merely makes the information filter obvious. The host compresses the probability of 99 unchosen doors into the single alternative he is allowed to leave closed.

Common reasoning mistakes

Treating survival as random

The alternative door did not remain by accident. The host’s knowledge determined which losing door disappeared.

Confusing a probability with a guarantee

A switching player still loses whenever the first pick is the car. The 2/3 result describes repeated games, not certainty in one round.

Forgetting to define the host

If the host can reveal the car, skip the offer, or choose strategically, the standard calculation may fail. State the rules before calculating.

Updating the first door without evidence

The host’s reveal gives no new reason to believe the first, blindly selected door was better than its original 1/3 chance.

What this puzzle teaches beyond game shows

The Monty Hall problem is a compact lesson about conditional probability, selection effects, and transparent models. It pairs naturally with base-rate neglect, where people overlook prior probabilities, and Simpson’s paradox, where aggregation changes the apparent result. Fermi estimation offers another way to strengthen quantitative judgment, while this guide to everyday mathematics provides a broader learning path.

The practical habit is simple: when new evidence appears, ask who selected it, what alternatives could have appeared, and which rules shaped the observation. That question matters in experiments, medical screening, hiring data, recommendation systems, and many other decisions.

Quick quiz with explanations

  1. What is the probability that the first pick is the car?
    Answer: 1/3, because the car is equally likely to occupy any of three doors.
  2. When does switching win?
    Answer: whenever the first pick is a goat, an event with probability 2/3.
  3. Why does the reveal carry information?
    Answer: the host knows the car location and deliberately avoids it.
  4. Does switching guarantee a win?
    Answer: no. It loses in the 1/3 of rounds where the first pick is correct.
  5. What is the switching probability with 100 doors under the same rules?
    Answer: 99/100.
  6. Would a random, uninformed door opening be the same experiment?
    Answer: no, because that person might reveal the car.
  7. Why might 30 simulated rounds differ from a 2/3 win rate?
    Answer: random variation is noticeable in small samples.
  8. What key error produces the fifty-fifty answer?
    Answer: counting the remaining doors while ignoring how the host selected the revealed door.

Sources and further study

Conclusion

The puzzle becomes straightforward once the host’s policy is included in the model. Staying wins only when the first guess was right, with probability 1/3. Switching wins whenever that guess was wrong, with probability 2/3. Under the standard rules, switch doors.

For an Indonesian companion lesson with a different simulation and examples, read Masalah Monty Hall: Mengapa Beralih Pintu Menang 2/3?.

cleverlysmart.com

Learn More →

Leave a Reply

Your email address will not be published. Required fields are marked *