Is a Coin Biased After 550 Heads in 1,000 Flips?

Reported in interviews at Google Accenture

Problem statement

A team wants to know whether a coin used in a game is fair. They model each flip as independent with the same probability $p$ of heads and test $H_0: p = 0.5$ against $H_1: p \ne 0.5$ at the 5% significance level. In a first experiment the coin is flipped 1,000 times and shows 550 heads. How many standard deviations above the fair-coin expectation is that? Then assess the evidence and plan the sample size for a follow-up test.

Show hint

Write down what the count of heads would look like if the coin were fair: its mean and its spread.

Relevant roles: Data Scientist and Applied Scientist.

Related free problems