# Simulate the Monty Hall Problem [closed]

I've never been able to wrap my head around the Monty Hall problem. Here's the premise:

Suppose you're on a game show, and you're given the choice of three doors: Behind one door is a car; behind the others, goats. You pick a door, say No. 1, and the host, who knows what's behind the doors, opens another door, say No. 3, which has a goat. He then says to you, "Do you want to pick door No. 2?" Is it to your advantage to switch your choice?

Run 10,000 simulations. Output the win percentage of switching. For example:

> 66.66733%


## closed as unclear what you're asking by ericw31415, DJMcMayhem♦, Blue, FryAmTheEggman, RɪᴋᴇʀMay 16 '16 at 14:55

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

• The host knows where the car is, so he never opens a door with the car behind it – bendytree Sep 30 '13 at 21:19
• As the challenge stands, it's hard to argue that it's not perfectly legal to just have 10,000 runs of the random-number generator and adds up which ones land below 0.6666. You might say that it's not really simulating the Monty Hall problem. But if it produces the exact same output, then what's really missing? – breadbox Oct 1 '13 at 0:32
• An interesting historical perspective (on the problem, not the golfing): Which Door Has the Cadillac?. – Cary Swoveland Oct 1 '13 at 22:29
• The challenge is still a little ambiguous with regards to what assumptions we're allowed to make. For example, some of the answers are probably saving a significant amount of code by assuming the car will be behind the same door every time or the player picks the same door every time. – Iszi Oct 25 '13 at 19:09
• Enumerating all the possibilities is at least as illuminating as generating them randomly... – Daniel Cristofani Nov 20 '13 at 22:03

## JavaScript 52

for(i=s=0;i++<1e4;)s+=Math.random()<2/3;alert(s/100)


Doors are 1:[0,1/3), 2:[1/3,2/3), 3:[2/3, 1)

Assume the prize is always in door 3. If the guest picks doors 1 or 2, which is range [0,2/3), and they switch, they've won the prize.

• Looks like CoffeeScript allows us to shave off a single character: i=s=0;s+=Math.random()<2/3while i++<1e4;alert s/100 – Kerrick Oct 5 '13 at 4:55

100%~+/2>?1e4$3  It picks a random door—let's label these 0, 1, or 2 where 2 is the door with the car—and computes the benefit of switching based on this logic: • If the player picked door 0, the host will open door 1. Switching will earn the player a new car (1). • If the player picked door 1, the host will open door 0. Switching will earn the player a new car (1). • If the player picked door 2, the winning door, the host will open either door 0 or door 1. Either way, if the player switches, he or she will find a goat (0). It then computes the result as the sum of the previous array, divided by 100. I'm pretty shaky with J so I'm sure this could be improved further. ## R 115 100 The pseudo-simulation answer is 23 characters long: sum(runif(1e4)>1/3)/100  but here is an actual simulation: D=1:3 S=function(x)x[sample(length(x),1)] sum(replicate(1e4,{ C=S(D) P=S(D) H=S(D[-c(C,P)]) F=D[-c(P,H)] C==F }))/100  1. D are the possible doors 2. S is a function to randomly select one item from a vector 3. C is the the door with the car (random among D) 4. P is the the door picked by the player (random among D) 5. H is the door picked by the host (random among D minus C and P) 6. F is the final door picked by the player (deterministic: D minus P and H) 7. success is measured by C==F. returns: [1] 66.731 Edit I can save a few characters by not assigning to variables and assuming without loss of generality that C==1: D=1:3;S=function(x)x[sample(length(x),1)];sum(replicate(1e4,{P=S(D);1==D[-c(P,S(D[-c(1,P)]))]}))/100  ## Perl, 9889837572 71 characters Here's a serious answer that actually runs the simulation: sub c{($==rand 2)-"@_"?$=:&c}$n+=$%==c c($%=rand 3)for 1..1e4;say$n/100  In each loop iteration, the player's initial choice is always door #2. First the door with the car is stored in $%, then a different door is selected for Monty Hall to expose. If the remaining door is equal to $%, the round is won. (Perl puncutation variables $% and $= are used because they do integer truncation for free.) ## Powershell - 168131125 115 Golfed code: nal g Random;1..1e4|%{$C=g 3;$P=g 3;$T+=$C-eq(0..2|?{$_-ne$P-and$_-ne(0..2|?{$_-ne$P-and$_-ne$C}|g)})};"$($T/100)%"


Changes from Original:
Trimmed 53 characters off the original script with some modifications.

• Removed spaces and parenthesis where PowerShell is forgiving of it.
• Used a ForEach-Object loop, via the % alias, instead of while.
• Used number ranges (e.g.: 0..2) instead of explicitly defined arrays.
• Removed write from the last command - turns out I don't need it after all.
• Flipped the expression for the host's choice around to use the shorter pipelining syntax.
• Replaced 10000 with 1e4.
• Took Joey's suggestion and omitted Get- from Get-Random. (Note: This modification significantly bloats the run time. On my system it jumped from about 20 seconds to nearly a half-hour per run!)
• Used Rynant's trick of doing $T+=... instead of if(...){$T++}.

Some notes:

This script is intended to be as concise as possible, while also being as thorough a simulation of the Monty Hall scenario as possible. It makes no assumptions as to where the car will be, or which door the player will choose first. Assumptions are not even made for which specific door the host will choose in any given scenario. The only remaining assumptions are those which are actually stated in the Monty Hall problem:

• The host will choose a door that the player did not pick first, which does not contain the car.
• If the player picked the door with the car first, that means there are two possible choices for the host.
• The player's final choice will be neither his initial choice nor the host's choice.

# Setup a single-character alias for Random, to save characters later.
# Note: Script will run a lot (about 500 times) faster if you use Get-Random here.
# Seriously, as it currently is, this script will take about a half-hour or more to run.
# With Get-Random, it'll take less than a minute.
nal g Random;

# Run a Monty Hall simulation for each number from 1 to 10,000 (1e4).
1..1e4|%{

# Set car location ($C) and player's first pick ($P) to random picks from a pool of 3.
# Used in this way, Random chooses from 0..2.
$C=g 3;$P=g 3;

# Increment win total ($T) if the car is behind the door the player finally chooses. # (Player's final choice represented by nested script.)$T+=$C-eq( # Filter the doors (0..2) to determine player's final choice. 0..2|?{ # Player's final choice will be neither their original choice, nor the host's pick. # (Host's pick represented by nested script.)$_-ne$P-and$_-ne(

# Filter the doors to determine host's pick.
0..2|?{

# Host picks from door(s) which do not contain the car and were not originally picked by the player.
$_-ne$P-and$_-ne$C

# Send filtered doors to Random for host's pick.
}|g
)
}
)
};

# After all simulations are complete, output overall win percentage.
"$($T/100)%"

# Variable & alias cleanup. Not included in golfed script.
rv C,P,T;ri alias:g


I've run this script several times and it consistently outputs results very near to two-thirds probability. Some samples:

(As above)

• 67.02%

(Using Get-Random as the alias definition, instead of just Random)

• 66.92%
• 67.71%
• 66.6%
• 66.88%
• 66.68%
• 66.16%
• 66.96%
• 66.7%
• 65.96%
• 66.87%

# Ruby 4840 38

My code doesn't make any assumptions about which door the prize will always be behind or which door the player will always open. Instead, I focused on what causes the player to lose. As per the Wikipedia article:

[...] 2/3 of the time, the initial choice of the player is a door hiding a goat. When that is the case, the host is forced to open the other goat door [...] "Switching" only fails to give the car when the player picks the "right" door (the door hiding the car) to begin with.

So to simulate this (instead of using fixed values), I modeled it as so:

• the show randomly picks 1 of the 3 doors to hide the prize behind
• the player then randomly picks 1 of the 3 doors as his first choice
• the player always switches, so if his first choice was the same as the show's choice, he loses

The code v1:

w=0;10000.times{w+=rand(3)==rand(3)?0:1};p w/1e2


The code v3 (thanks to steenslag and Iszi!):

p (1..1e4).count{rand(3)!=rand(3)}/1e2


Some sample return values:

• 66.44
• 66.98
• 66.33
• 67.2
• 65.7
• p (1..10000).count{rand(3)!=rand(3)}/1e2 saves some chars. – steenslag Dec 6 '13 at 20:15
• @steenslag Ah, indeed it does! Thank you! =) – Jonathan Hefner Dec 6 '13 at 21:28
• Does Ruby not allow shortcutting powers of 10? E.g: 1e4 for 10000? – Iszi Dec 6 '13 at 21:49
• @Iszi It does, but scientific notation yields a float, so it can't always be substituted. However, it is a viable substitution in v2, saving 2 more chars! – Jonathan Hefner Dec 6 '13 at 22:22

# Mathematica 42

Count[RandomReal[1,10^4],x_/;x>1/3]/100// N


66.79

PowerShell, 90

$i=0 1..10000|%{$g=random 3;$n=0..2-ne$g;$g=($n,2)[!!($n-eq2)]|random;$i+=$g-eq2}$i/10000


Commented:

# Winning door is always #2

$i=0 # Run simulation 1,000 times 1..10000|%{ # Guess a random door$g=random 3

# Get the doors Not guessed
$n=0..2-ne$g

# Of the doors not guessed, if either is the
# winning door, switch to that door.
# Else, switch to a random door.
$g=($n,2)[!!($n-eq2)]|random # Increment$i if
$i+=$g-eq2}

# Result
$i/10000  • This does not output a winning percentage in the format specified in the question. Also, save a couple characters by using 1e4 instead of 10000. – Iszi Nov 20 '13 at 20:44 ## C, 101 95 float c,s,t,i;main(){for(;i<1e5;i++,c=rand()%3,s=rand()%3)i>5e4&c!=s?t++:t;printf("%f",t/5e4);}  That's for the actual simulation. For some cheaty rule-bending code, it's only 71 65 59: p,i;main(){for(;i<1e5;rand()%5>1?i++:p++)printf("%f",p/1e5);}  I didn't do srand() because the rules didn't say I had to. Also, the cheaty version prints out about 30,000 extra numbers because it saves a character. I'm probably missing quite a few tricks, but I did the best I could. • Global variables are guaranteed to be zero on startup. Move your variable declarations out of main and you can drop the =0 initializations. – breadbox Nov 20 '13 at 19:58 Python 2: 72 66 64 from random import* i=10000 exec"i-=randint(0,2)&1;"*i print.01*i  Example output: 66.49 • You can save a few characters by using exec"i-=randint(0,2)&1;"*i instead of the for loop. – WolframH Oct 29 '13 at 22:09 • @WolframH Thanks, I'll update it now. – Rees Oct 29 '13 at 23:25 • Also, use print.01*i instead of print i/100.. – WolframH Oct 31 '13 at 0:32 • Nice solution but you're missing a semicolon. – Daniel Lubarov Nov 13 '13 at 22:07 • Very true. Updating now... – Rees Dec 6 '13 at 18:26 Fish - 46 43 This is using the same assumptions that Tristin made: aa*v>1-:?v$aa*,n;
v*a<$v+1$x! <
>a*0^<  $<  The down direction on x represents you initially picking the correct door, left and right are the cases that you picked a different door, and up is nothing, and will roll again. Originally, I initialized 10000 with "dd"*, but "dd" had to all be on the same line, and I wasted some whitespace. By snaking aa*a*a* I was able to remove a column, and ultimately 3 characters. There's a little bit of whitespace left that I haven't been able to get rid of, I think this is pretty good though! PHP 140 But i think that this is not working right. Any tip? I'm getting values from 49 to 50. $v=0;//victorys
for($i=0;$i<1e4;$i++){ //while the selection of the host ($r) equals the player selection or the car
//h=removed by host, p=player, c=car
while(in_array($h=rand(1,3),[$p=rand(1,3),$c=rand(1,3)])){} ($p!=$c)?$v+=1:0; //if the player changes the selection
}
echo ($v/1e4)*100;  • "If the player changes the selection"? – Timtech Nov 28 '13 at 16:33 • Sorry my English is not good. I mean, first i do a While trying to get aceptable values. Because the "host" can't remove a door containing the car OR the door that you choose. Then i have$p (players choice) and $c (where the car is). The OP said that you must take the percentage of winning when you switch, so i only count the result as a "victory" when$p!=\$c (you switch your choice to the other door and you win). – Carlos Goce Nov 28 '13 at 16:59

## Game Maker Language, 19 (51 w/ loop)

show_message(200/3)


It outputs 66.67! This is the correct probability ;)

The serious-mode code, 51 characters:

repeat(10000)p+=(random(1)<2/3);show_message(p/100)


Make sure to compile with treat all uninitialized variables as 0.

The oldest code, 59 characters:

for(i=0;i<10000;i+=1)p+=(random(1)<2/3);show_message(p/100)


Again, make sure to compile with treat all uninitialized variables as 0.

The output was 66.23