Open Side Menu Go to the Top
Register
Fun puzzle: How to make an unfair coin fair Fun puzzle: How to make an unfair coin fair

11-24-2010 , 08:02 PM
I read about this old, elegant puzzle, having forgotten about it nearly a decade ago and thought some here might enjoy it. It's trivial to google for the answer so I urge you to resist that temptation. PLEASE DON'T POST THE ANSWER (not like I can stop you), but feel free to brag if you figured it out.

On to the problem:

Let's assume you are handed a coin, and told that it is biased, but you are told neither in which direction the coin is biased (heads or tails) nor how strong the bias is (70%/30%, 51%/49%...). Your task - devise a procedure, using only the biased coin, that results in either HEADS or TAILS with a probability of exactly 50% for either outcome. You may use as many coin tosses as you want to generate a HEADS or TAILS.

Cheers,
bachfan

Don't
scroll
down
in
case
someone
posted
the
answer.

And
don't
post
the
answer
if
you
already
know
it.

Fa
la
la
la
la
la
la
la
la.

Happy
holidays
everyone.
Fun puzzle: How to make an unfair coin fair Quote
11-25-2010 , 09:46 AM
Spoiler:

I've tried to figure this one out... it must be something to do with a sequence of tosses coming up a certain way. I'm mucking about with binary numbers at the moment to try and see if that gives a solution.
Fun puzzle: How to make an unfair coin fair Quote
11-25-2010 , 09:50 AM
Solution (I think):

Spoiler:
Hang on I think I have it. Flip the coin twice. If it comes up the same twice then spin again... if it comes up different then the second toss counts. I think that this is equally likely. Lets assume the probablity of head is h and tail is 1-h
> Head Head has probability h^2 - spin again
> Tail Tail has probability (1-h)^2 - spind again
> Head Tail has probability h(1-h) - counts as TAIL with probability h-h^2
> Tail Head has probability (1-h)h - counts as HEAD with probability h-h^2

h-h^2 is always positive where h isn't 1 or zero.

...that's it.

Nice problem.


Fun puzzle: How to make an unfair coin fair Quote
11-25-2010 , 12:39 PM
von neumann corrector.
Fun puzzle: How to make an unfair coin fair Quote
11-25-2010 , 12:50 PM
Nice job hep!
Fun puzzle: How to make an unfair coin fair Quote
11-25-2010 , 12:52 PM
Quote:
Originally Posted by Pyromantha
von neumann corrector.
Indeed. Old tricks are the best tricks.
Fun puzzle: How to make an unfair coin fair Quote
11-25-2010 , 04:12 PM
I thought of that and thought there must be a more elegant solution.
Fun puzzle: How to make an unfair coin fair Quote
11-25-2010 , 04:18 PM
Try the one I asked spadebidder yesterday, it's more interesting:

You have a run calculator for which you can enter 3 numbers m,n,p and it will tell you the probability of at least 1 run of length at least m of events with probability p in n trials. For example, if n=100, m=5, p=0.5 it will tell you the probability that you will get at least 5 consecutive heads in 100 flips of a fair coin. What numbers would you enter to compute the probability of 5 consecutive heads OR tails in 100 coin flips? Note that these are not independent events.
Fun puzzle: How to make an unfair coin fair Quote
11-25-2010 , 04:25 PM
Quote:
Originally Posted by hepzebah
Spoiler:

I've tried to figure this one out... it must be something to do with a sequence of tosses coming up a certain way. I'm mucking about with binary numbers at the moment to try and see if that gives a solution.
That's how you generate an arbitrary probability given a fair coin.
Fun puzzle: How to make an unfair coin fair Quote
11-26-2010 , 10:03 AM
Quote:
Originally Posted by BruceZ
Try the one I asked spadebidder yesterday, it's more interesting:

You have a run calculator for which you can enter 3 numbers m,n,p and it will tell you the probability of at least 1 run of length at least m of events with probability p in n trials. For example, if n=100, m=5, p=0.5 it will tell you the probability that you will get at least 5 consecutive heads in 100 flips of a fair coin. What numbers would you enter to compute the probability of 5 consecutive heads OR tails in 100 coin flips? Note that these are not independent events.
My first 3 guesses are, without any checking or real thinking....
> Double the flips, otherwise unchanged
> 1 minus the probability squared, otherwise unchanged
> m = 4, otherwise unchanged

...will have a think now.
Fun puzzle: How to make an unfair coin fair Quote
11-27-2010 , 11:27 AM
Spoiler:
Ignorance is key!
You throw twice. If it comes heads heads or tails tails you ignore it.
1.heads 2.tails we call heads
1.tails 2.heads we call tails
since they are equally likely it's 50%.

Nice one bachfan, took me a while.
Fun puzzle: How to make an unfair coin fair Quote
11-27-2010 , 04:57 PM
Quote:
Originally Posted by hepzebah
My first 3 guesses are, without any checking or real thinking....
> Double the flips, otherwise unchanged
> 1 minus the probability squared, otherwise unchanged
> m = 4, otherwise unchanged

...will have a think now.
One of these guesses is very close.
Fun puzzle: How to make an unfair coin fair Quote
12-01-2010 , 05:22 PM
Seriously, nobody?

Answer:
Spoiler:

Instead of heads/tails, consider same/different from the previous flip which also have probability 0.5. For 5 consecutive heads or tails in 100 flips, we need 4 consecutive "sames" in 99 flips while the 1st flip can be anything. So m=4, n=99, p=0.5.
Fun puzzle: How to make an unfair coin fair Quote
12-02-2010 , 05:45 AM
Yeah I got that Bruce, with a little help from your post. I used the inspection method (i.e. guesswork) to solve.
Fun puzzle: How to make an unfair coin fair Quote
12-03-2010 , 04:16 AM
Quote:
Originally Posted by BruceZ
Seriously, nobody?

Answer:
Nice puzzle. Wish I'd been on 2+2 to mull it over before you spoiled it (I have little self-control sometimes!)
Fun puzzle: How to make an unfair coin fair Quote
12-04-2010 , 06:26 AM
Sooooo Eeezeeeeee, do I get to put another star on my wizard hat? Oh forgot my wizard hat has dollar signs...
Fun puzzle: How to make an unfair coin fair Quote

      
m