Math - Probability

The Three Doors Problem (Monty Hall) Explained Simply

Written by Viktor Jonas · Reading time about 5 minutes

Three doors: behind one is a car, behind each of the others a goat. You pick a door. The host — who knows where the car is — opens a different door with a goat behind it and asks whether you want to switch. Should you? Yes. Always.

stay: 1/3  ·  switch: 2/3 Switching doubles your chance of winning
Key terms

Conditional probability — the probability given some extra information.
Extra information — here: the host knows where the car is and acts accordingly.
Complementary event — “picked wrongly” has probability 2/3.
Monty Hall problem — the name the puzzle is known by in English, after the game-show host Monty Hall.

Why the chance is not 50:50

The crucial point: the host does not open a random door. He knows where the car is and always opens a door with a goat. So his action carries information — and that information is not shared equally between the two remaining doors.

All three cases played through (car behind door 1)
You pickHost opensstayswitch
Door 1 (car)Door 2 or 3winsloses
Door 2 (goat)Door 3loseswins
Door 3 (goat)Door 2loseswins

In two out of three cases, switching wins. No more proof is needed — you only have to write down all the cases completely.

Your pick 1/3 Switch door 2/3

The trick with a hundred doors

Imagine a hundred doors. You choose one — your chance is 1%. The host then opens 98 goat doors and leaves exactly one other door closed. Would you stay with your first choice now? Hardly anyone would. With three doors it is exactly the same situation, just less obvious.

💬 Mia asks Grandpa Theo

Mia

But in the end there are only two doors left. So it's 50:50!

Grandpa Theo

Two possibilities does not mean equally likely. Your first pick was right with probability 1/3 — opening a door does not change that.

Mia

And the other 2/3?

Grandpa Theo

They now sit entirely on the one remaining door. The host has sorted out the bad possibility for you.

Common mistakes

“Two doors, so 50:50.”
The same number of options does not mean the same probability.
Overlooking what the host knows
If he opens a door at random and happens to hit a goat, it really is 50:50 — the rule he follows is decisive.
Looking at only one case
Only a complete case-by-case analysis shows the result.
Trusting intuition
Even mathematicians have been wrong about this in public.
Practice problem

How does the result change if the host opens the second door purely at random and happens to hit a goat? Explain.

Show solution

Then the chance of winning is 50:50, and switching gains nothing.

The reason: in this case the host could also have revealed the car. That he shows a goat is now itself a random outcome and gives no additional information about the remaining doors. In the original problem, by contrast, he follows a fixed rule — and exactly this rule transfers the 2/3 to the one remaining door. It is not the number of doors that decides, but the behaviour of the host.

Why our gut feeling fails here

The three doors problem became famous because even maths professors at first considered the correct solution wrong. It is worth a closer look, because behind it lies a thinking error that reaches far beyond this puzzle.

Our brain likes to treat number and probability as the same thing. Two remaining doors feel like two equally good possibilities. In fact, though, probability does not depend on how many possibilities are left, but on how the information came about. The host does not act randomly: he knows where the car is, and he is guaranteed to open a losing door. So in every round he gives away information — and that information lands entirely on the door he did not open.

All three cases played through once

The surest way to be convinced is to write down every possible case. Suppose you always pick door 1.

Case 1: The car is behind door 1. The host opens door 2 or 3, both are losing doors. Switching loses.
Case 2: The car is behind door 2. The host must open door 3, because you picked door 1 and door 2 hides the car. Switching wins.
Case 3: The car is behind door 3. The host must open door 2. Switching wins.

Three cases, all equally likely, and in two of them switching wins. That is exactly the two thirds. If you would rather try it out: with two people and three playing cards you can play it through thirty times in ten minutes, and the result visibly settles near two to one.

By the way, this error is not stupidity but a pattern. People regularly misjudge probabilities as soon as extra information comes into play — especially when it is not obvious whether that information arose by chance or on purpose. This distinction is exactly the core of conditional probability, and the three doors problem is its most memorable example.

🎥 Explained in under two minutes – as a Short on our YouTube channel.

Watch the Short

In 1990 Marilyn vos Savant published the correct solution in a newspaper column. About 10,000 letters followed, many from universities, accusing her of a mistake in her calculation. She was right.

Practice: five questions, instant feedback

Click an answer. You see straight away whether it is right — and why.

  1. What is your chance of winning if you stay with your first door?

    1/3. Your first pick was right one time in three, and opening a goat door changes nothing about that — the host could have opened such a door in every case.

  2. Why is the situation at the end not 50:50?

    His knowledge is the heart of it. Because he never reveals the car, his action carries information — and all of it collects in the one door he leaves closed.

  3. You pick door 2, the car is behind door 1. Which door does the host open?

    Door 3. He may not open your door and he may not show the car, so only one option is left. This is exactly the case in which switching wins.

  4. Now the host opens one of the other two doors completely at random and happens to hit a goat. What holds then?

    Then it really is 50:50. His action tells you nothing any more, because he might have revealed the car. What decides the answer is not the number of doors but the rule he opens by.

  5. With 100 doors you pick one, the host opens 98 goat doors. How good is switching?

    99/100. Your first pick holds 1/100. The remaining 99/100 does not spread out — it piles up on the single door the host deliberately left closed.

0 of 5 answered.

Exam questions with model answers

Try it yourself first, then open the answer.

Using a complete case analysis, show why switching wins with probability 2/3.

Put the car behind door 1 without loss of generality and go through the first pick, which falls on each of the three doors with probability 1/3.

Pick door 1: the host opens door 2 or 3, switching leads to a goat — lost. Pick door 2: he must open door 3, switching leads to door 1 — won. Pick door 3: he must open door 2, switching leads to door 1 — won.

Two of the three equally likely cases win when you switch, so P = 2/3. For full marks you also have to say why the three cases are equally likely and why the host has no choice in two of them.

Derive the same result with Bayes' theorem.

Say you picked door 1 and the host opened door 3. Events: A₁, A₂, A₃ mean „car behind door 1, 2 or 3“, each with P = 1/3. M₃ means „host opens door 3“.

The conditional probabilities: P(M₃ | A₁) = 1/2, because he may choose between doors 2 and 3. P(M₃ | A₂) = 1, because door 1 is picked and door 2 hides the car. P(M₃ | A₃) = 0, he does not show the car.

So P(M₃) = 1/3 · 1/2 + 1/3 · 1 + 0 = 1/2. Substituting: P(A₁ | M₃) = (1/3 · 1/2) / (1/2) = 1/3 and P(A₂ | M₃) = (1/3 · 1) / (1/2) = 2/3. Staying wins with 1/3, switching with 2/3 — the same result as the case analysis.

Explain why so many people guess 50:50, and name the condition that would actually make it 50:50.

The usual mistake is to count doors instead of reading the rule. Two doors are left, so the gut says one in two. That ignores the fact that the host acted under a constraint: he knows the arrangement and never opens the car.

It really is 50:50 only if the host opens at random and merely happens to hit a goat. Then his action carries no information, and both remaining doors hold 1/2. The number of doors is the same in both scenarios — only the rule differs, and the rule is what decides.

Sources and further reading

Everything on this page can be checked here.

← Back to overview