Monty Hall Problem

Everyone

Every game played here, added up.

The claim: it is better to switch

The maths makes one prediction, and it makes it before anyone plays. In a three door game your first pick is wrong two times in three. That is 66.67%, exactly. So switching wins twice as often as staying, and it does not matter who is playing, how well, or whether they believe it.

That does not mean switching wins every time. Any one game can go either way, and someone who switches on all 100 still drives home a goat about a third of the time. The claim is about the long run, where the gap is not close. Everything below is that prediction put to the test by people clicking doors.

What the games produced

Across 500 finished games, switching would have won

66.60%

of the time. Theory says 66.67%, so these games land 0.07 percentage points away. That is the same figure as how often the first pick was not the car, because those are the same games.

Finished runs5
Unfinished1
Completion rate83.3%
Still playing0

One run proves nothing. All of them make a shape.

A single hundred-game run can land anywhere: some people get 58, some get 76. That is why one run never settles an argument. But stack every finished run side by side and the scatter is not random noise, it is a shape, and the shape was known in advance. The curve is Binomial(100, 2/3), centred on 67 with a standard deviation of 4.7. It was not fitted to the bars. It is what the maths said the bars would look like.

2 0 0: 01: 02: 03: 04: 05: 06: 07: 08: 09: 010: 011: 012: 013: 014: 015: 016: 017: 018: 019: 020: 021: 022: 023: 024: 025: 026: 027: 028: 029: 030: 031: 032: 033: 034: 035: 036: 037: 038: 039: 040: 041: 042: 043: 044: 045: 046: 047: 048: 049: 050: 051: 052: 053: 054: 055: 056: 057: 058: 059: 060: 061: 062: 063: 164: 265: 066: 067: 068: 069: 170: 071: 072: 073: 174: 075: 076: 077: 078: 079: 080: 081: 082: 083: 084: 085: 086: 087: 088: 089: 090: 091: 092: 093: 094: 095: 096: 097: 098: 099: 0100: 0 67 expected 0102030405060708090100
Finished runs What theory predicted

Even the people who walked away

1 runs stopped early, after 13 games between them. A run counts as left behind once it sits untouched for 30 minutes, and runs still being played are not counted here. Their games are shorter and their results noisier, so they are plotted as a rate in 5 point steps rather than a raw count. The puzzle does not care whether anyone finished: the first picks they did make came out wrong at the same rate.

1 0 0%: 05%: 010%: 015%: 020%: 025%: 030%: 035%: 040%: 045%: 050%: 155%: 060%: 065%: 070%: 075%: 080%: 085%: 090%: 095%: 0 two thirds 0%20%40%60%80%100%
Unfinished runs

Pooled across every unfinished game

66.67% expected observed 53.85%

The luckiest run so far

Best run

64 cars

The most cars won in a single finished run, out of 100. Switching every time averages 67, so anything well above that is a good run rather than a better method. There is no technique hiding here.

Add your own

Every figure on this page is the sum of people who sat down and clicked through a hundred games. A run takes about ten minutes.

Play a hundred games

Recomputed every 30 seconds. Last at 13:59:08 UTC.