The answer is the expected proportion of Heads and Tails is 50:50.
A plausible intuition is that the strategy described in the problem with result in more Heads being observed than Tails, but this intuition is misleading
Each coin flip is independent and over a large number of trials Heads and Tails will each turn up with about the same frequency, and no strategy we might decide to follow can change this ratio.
I wrote a Python program to simulate the problem and sure enough if we run it Heads and Tails each appear in with about the same frequency.
import random
H = 0
for i in range(1000000):
throw = random.choice(['H','T'])
if throw != 'T':
H = H + 1
while throw != 'T':
throw = random.choice(['H','T'])
if throw != 'T':
H = H + 1
print('Heads per 1000000 Tails', H)
Comments
New comment
Hi, Richard
I run the python program and got the output,
'Heads per 1000000 Tails 999965'
This program looks doing Monte Carlo Simulation.
What provability distribution does the problem follow?
New comment
I find it interesting to approach this problem from the other direction. Instead of considering what happens for very many people, we can consider the expectation value for the number of heads for a single person flipping one coin.
There is a 50% probability of the first coin flip being tails, and so of there being 0 heads. There is then a 25% probability of obtaining heads once, since there is a 50% probability of the first flip being heads and 50% thereafter of the second flip being tails. Proceeding in this way, we have a probability of obtaining heads times, so that the expectation value is the infinite series
.
This series can be evaluated similarly to Example 2 under the heading Taylor Series in this document. The result is that we expect to see 1 coin come up heads on average, which is equal to the number of tails we always have.
Hence it does not matter how many or how few people are flipping coins—the expected proportion of heads to tails for a single person is 1:1, and each person is flipping their own coin independently, so this is always the case.
New comment
Hi, Richard and Steven.
So, that’s why the output of the program 999965 is close to 1000000, thank you. Maybe, this probability distribution is the sum of multiplication of Bernoulli trial probability,
P=P(1)*P(0)+P(1)*P(1)*P(0)+P(1)*P(1)*P(1)*P(0)+…
Where,
1:the event of head
0:the event of tail
P(*): probability of the event *
P:The expected proportion of heads to tails for a single person
New comment
p.s.
Sorry. Maybe, the number of heads for single person may be
N=1*P(1)*P(0)+2*P(1)* P(1)*P(0)+ 3*P(1)* P(1) * P(1)*P(0)+…