Mathwords logoMathwords

St. Petersburg Paradox — Definition, Formula & Example

The St. Petersburg Paradox is a coin-flipping game whose expected payout is infinite. Even so, almost no one would pay more than a few dollars to play it.

Flip a fair coin until the first head appears. If the first head occurs on flip kk, the payout is 2k2^k dollars, which happens with probability (1/2)k(1/2)^k. The expected payout ∑k=1∞(1/2)k⋅2k\sum_{k=1}^{\infty} (1/2)^k \cdot 2^k diverges. The paradox is the conflict between this infinite expected value and the small finite price a reasonable person would pay. It shows that maximizing expected monetary value cannot be the whole rule for rational decisions.

Key Formula

E[X]=∑k=1∞(12)k⋅2k=∑k=1∞1=1+1+1+⋯=∞E[X] = \sum_{k=1}^{\infty} \left(\tfrac{1}{2}\right)^k \cdot 2^k = \sum_{k=1}^{\infty} 1 = 1 + 1 + 1 + \cdots = \infty
Where:
  • XX = The payout of the game, in dollars
  • kk = The flip on which the first head appears (k = 1, 2, 3, ...)
  • (1/2)k(1/2)^k = Probability that the first head lands on flip k
  • 2k2^k = Payout in dollars when the first head lands on flip k

Worked Example

Problem: A real casino has limited money. It caps the payout at $1,048,576, which equals 2202^{20} dollars. If the first head comes after flip 20, you still get only $1,048,576. What is the expected payout of this capped game?
Step 1: For flips 1 through 20, the payout is uncapped. Each possible outcome contributes exactly $1 to the expected value, so these 20 outcomes contribute $20.
∑k=120(12)k⋅2k=∑k=1201=20\sum_{k=1}^{20} \left(\tfrac{1}{2}\right)^k \cdot 2^k = \sum_{k=1}^{20} 1 = 20
Step 2: Next, take the case where the first 20 flips are all tails. This happens with probability (1/2)20(1/2)^{20}, and you receive the cap. This case adds $1.
(12)20⋅1,048,576=220220=1\left(\tfrac{1}{2}\right)^{20} \cdot 1{,}048{,}576 = \frac{2^{20}}{2^{20}} = 1
Step 3: Add the two parts to get the expected payout.
E[X]=20+1=21E[X] = 20 + 1 = 21
Answer: The expected payout is only $21. A cap above a million dollars cuts the 'infinite' value down to $21. All of the infinity comes from extremely rare, astronomically large payouts.

Why It Matters

Daniel Bernoulli proposed a resolution in 1738. He argued that people value the utility of money, not the dollar amount itself. With utility ln⁡(x)\ln(x), the expected utility is ln⁡2⋅∑k=1∞k/2k=ln⁡4\ln 2 \cdot \sum_{k=1}^{\infty} k/2^k = \ln 4. So a sure $4 gives the same expected utility as the game's random payout, making it worth only about $4 to a player with log utility and no other wealth. This idea of expected utility, with diminishing returns on extra money, is the foundation of risk aversion in economics, insurance pricing, and finance courses.

Common Mistakes

Mistake: Assuming an infinite expected value means you will probably win a huge amount.
Correction: Expected value is a long-run average, not a typical result. Half the time you win only $2. Three-quarters of the time you win $4 or less. Very rare enormous payouts are what push the average to infinity.

Related Terms

  • Bernoulli Trials — Each independent fair coin flip in the game
  • Experiment — Flipping until the first head is the random experiment
  • Event — First head on flip k defines each payout event
  • Expected Value — Its expected payout is the quantity that turns out infinite
  • Diverge — The payout series fails to settle on a finite sum