The Game Worth Infinity
The puzzle
A coin is tossed until it comes up tails. If the first tails appears on toss n, you are paid 2^(n-1) pounds: £1 if the very first toss is tails, £2 if the second, £4 if the third, and so on.
What is the expected payout, and what is the most you would pay to play?
Stuck?
One at a time, smallest first. Open only as many as you need — the whole point of the puzzle is the part you work out yourself.
Hint 1
Work out the contribution of each outcome: probability times payout.
Hint 2
Every single term contributes the same amount.
Show the answer Hide the answer
The expected payout is infinite - each term contributes 50p and there are infinitely many. Almost nobody would pay more than about £20, and that gap is the paradox.
How to get there
The first tails appears on toss n with probability 1/2^n, and the payout is 2^(n-1), so each term contributes (1/2^n) * 2^(n-1) = 1/2 pounds. The expected value is 1/2 + 1/2 + 1/2 + ... which diverges.
But a casino with finite money can only offer a finite game. Cap it at 20 tosses and the expectation is 20 * 1/2 = £10 - and the maximum payout is already over half a million. Cap it at 40 tosses and the expectation is only £20, while the top prize exceeds half a trillion.
That is the resolution. The expectation is driven entirely by outcomes of vanishing probability and astronomical payout, so it grows like the logarithm of the bank. Median winnings are £2 whatever the cap. Expected value is the right guide when you can repeat a gamble many times; for a one-off bet with a long tail it answers a question nobody asked.
The technique worth keeping
When an expectation is dominated by rare extreme outcomes, check it against the median and against the finite version of the game. A divergent expected value is a signal that the model has outrun the situation, not that the bet is good.