What Is the Monty Hall Problem and Why Our Intuition Gets It Wrong

Three Doors and a Car

The year is 1990. Columnist Marilyn vos Savant receives the following question in her "Ask Marilyn" column in Parade magazine: Suppose you are on a game show and you are given the choice of three doors. Behind one door is a car; behind the other two are goats. You pick a door — say, door number one. The host, who knows what is behind each door, opens another door — say, door number three — which reveals a goat. He then asks: do you want to switch to door number two, or stay with door number one?

Vos Savant answered that you should always switch — doing so gives you a 2/3 chance of winning the car, while staying with your original choice gives you only a 1/3 chance. The response was volcanic. Nearly 10,000 readers wrote in to dispute her answer, including many with PhD credentials in mathematics. Several wrote to inform her that she had made a basic probabilistic error. Almost all of them were wrong. The Monty Hall problem — named after the host of the American game show Let's Make a Deal — had claimed its victims once again.

Why Your Intuition Is Wrong

The intuitive reasoning goes like this: after the host opens a door, there are two doors left, one hiding a car and one hiding a goat. So the probability must be 50/50, and it does not matter whether you switch. This argument is wrong, but articulating why it is wrong requires careful attention to what the host's action actually tells you.

The key insight is that the host's behavior is not random. The host always opens a door with a goat, and the host never opens the door you chose. This constraint means the host's action carries information — information that changes the probabilities. When you initially picked your door, the probability of the car being there was 1/3. The probability of the car being behind one of the other two doors was 2/3. When the host opens one of those other doors to reveal a goat, the 2/3 probability does not split evenly between the remaining doors — it collapses entirely onto the one unopened door the host did not choose. Switching captures that 2/3 probability.

Working Through the Cases

The clearest way to see the correct answer is to enumerate all possible cases. Suppose the car is always behind door 1 (since the logic is the same regardless of where it actually is, this simplifies the example). You can initially pick any of the three doors:

1. You pick door 1 (car): The host opens either door 2 or door 3 (both have goats). If you switch, you get a goat. Switching loses.
2. You pick door 2 (goat): The host must open door 3 (it is the only other goat-containing door not chosen by you). If you switch to door 1, you get the car. Switching wins.
3. You pick door 3 (goat): The host must open door 2. If you switch to door 1, you get the car. Switching wins.

In two out of three equally likely initial scenarios, switching wins. In one out of three, staying wins. Switching gives you a 2/3 probability of winning. The calculation confirms the intuition-defying answer: always switch.

The Role of the Host's Knowledge

What if the host did not know where the car was and opened a door at random? Then if the host happened to reveal a goat, the probabilities would indeed be 50/50 for the remaining two doors. This version is called the Monty Fall problem (or the Ignorant Monty version), and it illustrates why the host's knowledge is crucial.

When the host knows where the car is and deliberately avoids revealing it, the host's action carries encoded information. The doors you did not pick had a 2/3 combined probability of hiding the car; the host's action eliminates one of them while telling you (implicitly) that the remaining unchosen door is the one most likely to hide the car if your initial guess was wrong. This is a concrete example of how the source of information matters as much as the information itself — a principle central to Bayesian reasoning and information theory.

Why PhDs Got It Wrong

The fact that thousands of people with mathematical training initially rejected the correct answer is itself an important phenomenon. Paul Erdos, one of the most prolific mathematicians in history, reportedly refused to believe the correct answer until he was shown a computer simulation. What goes wrong?

Several cognitive biases contribute. The most important is the tendency to model the situation as a symmetric 50/50 choice once there are two options remaining, ignoring how those options came to be the remaining ones. This is related to the gambler's fallacy — the mistaken belief that the probability of an outcome is affected only by the current number of possibilities, not by the process that generated the situation. Additionally, humans are poor at tracking conditional probability — probability given that something else is true — and the Monty Hall problem requires precisely this tracking.

Simulations and Empirical Tests

The beautiful thing about the Monty Hall problem is that it is testable. Anyone can simulate it with playing cards, coin flips, or a simple computer program. When vos Savant encouraged readers to verify the answer empirically, many elementary school students ran classroom simulations and confirmed the 2/3 result. The mathematics can be verified in an afternoon with a deck of cards.

Computer simulations run millions of trials and consistently show: switching wins approximately 66.7% of the time; staying wins approximately 33.3% of the time. The discrepancy from 50/50 is not a matter of rounding or subtle statistical tricks — it is robust across any number of trials. For many people, only seeing the simulation results makes the counterintuitive answer click into place.

Generalizations of the Problem

The Monty Hall problem generalizes in interesting ways. Suppose there are 100 doors instead of three — 99 with goats, one with a car. You pick one door. The host then opens 98 doors, all revealing goats, leaving only your door and one other. Should you switch? In this version the correct answer feels much more intuitive: your initial pick had a 1/100 chance of being right. The host's dramatic elimination of 98 other doors concentrates the remaining 99/100 probability entirely on the one door the host did not open. Of course you should switch.

This generalization reveals the underlying logic: the more doors the host eliminates while respecting the constraint (never open the car, never open your door), the more information concentrates on the remaining unchosen door. With three doors, the effect is present but subtle enough to fool most people; with 100 doors, it is obvious. The mathematical structure is identical — only the scale changes.

Lessons Beyond the Puzzle

The Monty Hall problem teaches lessons that extend far beyond game shows. It illustrates that changing new circumstances can change optimal strategies even when the raw number of options stays the same. It demonstrates that our intuitions about probability are systematically biased in ways that careful mathematical analysis can correct. And it shows that even experts — including mathematicians — are not immune to these biases. Epistemic humility, combined with rigorous analysis and willingness to update when evidence contradicts intuition, is the proper response to problems like Monty Hall. The doors are always trying to fool you.