The Monty Hall Problem: When Intuition Says the Wrong Answer

In 1990 a reader wrote to Marilyn vos Savant's Parade magazine column with a game-show puzzle: three doors, a car behind one, goats behind two. You pick a door. The host, who knows what's where, opens a different door revealing a goat. Should you switch? Vos Savant answered: yes — switching wins two-thirds of the time. Then ten thousand letters arrived, many from university mathematicians, calling her everything short of innumerate. The professionals with doctorates were certain it was fifty-fifty. The columnist with the puzzle column was right, and the PhDs were wrong, and the whole episode has been embarrassing experts and delighting families ever since.

I retell it because it is the cleanest exhibit ever staged of the gap between feeling probable and being probable — and that gap is where students live every day. Our experimental probability guide shows what happens when trials defy expectations, and our probability rules guide builds the and-or machinery this puzzle quietly uses. This article stages the fight itself: your intuition versus the two-thirds truth, with a kitchen-table experiment as referee.

The setup and the seductive wrong answer

Fix the rules first, because every fake refutation changes them quietly. Three doors: one car, two goats. You choose one — say door 1, and suppose the car is actually behind door 3. The host, who knows, must open a remaining door with a goat: door 2 is forced (he cannot open your door, cannot reveal the car). Now two doors remain closed and you are offered the switch. Intuition counts two closed doors and reports fifty-fifty, the way it reports every binary-looking choice. That count ignores the host's knowledge, which already did half the work of eliminating for you.

The fifty-fifty feeling has a precise source worth naming: the equally-likely fallacy, the reflex that N remaining options means 1/N each. It holds for fresh randomness — a new coin flip is fifty-fifty no matter the history — but these doors are not fresh. Information entered between your pick and the offer (a knowing host removed a goat), and information redistributes probability the way wind redistributes leaves. Two closed doors can hold unequal chances when one of them survived a selection and the other merely sat there. Your first pick holds its original one-third; everything else pooled into the other door.

Why switching wins: the three worlds

List all three possible worlds — car behind 1, 2, or 3, each with probability one-third — and play the switch strategy in each. Car behind 1 (your pick): host opens either goat door, you switch away from the car and lose. Car behind 2: host is forced to open door 3, you switch to door 2 and win. Car behind 3: host forced to door 2, you switch to door 3 and win. Two winning worlds out of three. The table never lies because it contains every case instead of the one case your gut examined:

Still unconvinced? Scale it to one hundred doors. You pick one; the host, knowing, opens ninety-eight goat doors, leaving yours and one survivor. Staying still claims your first blind guess beat ninety-nine alternatives; switching claims the host's careful elimination points at the car. Nobody feels fifty-fifty about the hundred-door version — the intuition flips instantly — yet it is the same arithmetic wearing a bigger number. If the hundred doors convince you and the three doors don't, your beliefs contradict each other, and the table above is the referee that never blinks. Our probability trees guide draws exactly this case-splitting as branches, which is why tree people rarely stay confused past the second branch.

Prove it at the table in thirty rounds

Arguments end debates; tallies end wars. Deal three cards — one ace for the car, two number cards for goats — and play thirty rounds with the family, always switching, shuffling the ace position each time. Mark wins and losses in two columns where everyone watches. Around round twelve the room goes quiet, because the switch column pulls ahead and keeps pulling; by round thirty it holds roughly twenty wins against ten, and the fifty-fifty caucus has no furniture left. Thirty rounds take ten minutes and outperform thirty minutes of proof, because the hands learn what the head refused.

Then run the killer variation: play ten rounds staying every time. The stay column lands near three wins in ten — the one-third your original pick always owned, untouched by theater. Side by side, the two columns are the whole lesson in one glance: staying keeps your birthright third, switching harvests the redistributed two-thirds. Children who watched the columns grow never relapse into fifty-fifty on this problem, and they gain something larger — the lived experience that experiments outrank instincts, which is the entire scientific attitude in miniature. Our statistics overview explains why thirty rounds already speak clearly while three rounds prove nothing: small samples wobble, large ones confess.

The version where fifty-fifty is correct

Here is the twist that separates understanding from memorized doctrine: change one rule and the answer flips. Suppose the host does NOT know where the car is and opens a remaining door at random — and happens to reveal a goat. Now switching gains nothing, and fifty-fifty is exactly right. Why? Because the random host might have revealed the car instead (ending the game), so the worlds where play continues are symmetric: your door and the survivor each hold the car in half of the surviving worlds. The knowing host's forced choice funnels probability onto the survivor; the random host's lucky miss funnels nothing anywhere.

This is why mathematicians who answered vos Savant were not entirely hallucinating — many solved this random-host version in their heads while reading the knowing-host problem. The two scenarios sound identical in casual speech ("he opens a door, shows a goat") and differ only in the host's knowledge, which the puzzle states once and readers skip once. Teach both versions side by side and something precious happens: the child learns that probability lives in the rules, not in the objects. Same three doors, same goats, opposite answers — because the question was never about doors at all. It was always about what the host knew.

Why brains refuse — and where else they refuse

The stubbornness has structure worth studying, because the same three saboteurs ambush students everywhere. First, endowment: your door feels like property, and abandoning property feels like loss even when the trade is profitable — the same bias that keeps losing stocks in portfolios. Second, the action-omission asymmetry: losing by switching stings more than losing by staying, so caution masquerades as probability, and families should name this aloud whenever a child defends fifty-fifty with unusual heat. Third, the fresh-randomness reflex already covered: two options, therefore halves, history ignored.

Once named, the saboteurs show up all over mathematics, and spotting them becomes a transferable skill. The birthday problem's answer (twenty-three people, fifty-fifty shared birthday) feels impossible for the same fresh-randomness reason — pairs multiply invisibly, 253 of them, while intuition counts people instead of pairs. Gambler's fallacy runs the mirror image: five heads make tails feel due, though the coin holds no memory. Teach the trio together — Monty Hall, birthdays, streaks — as one lesson with three costumes: intuition counts the visible options and skips the hidden structure, and mathematics is the discipline of counting what intuition skipped.