What Is Ergodicity? (Taleb Definition)
Ergodicity refers to whether the time-averaged outcome for one individual matches the ensemble-averaged outcome across many individuals.
A system is ergodic when these two averages coincide — when what happens to you over time is representative of what you'd expect from looking at a cross-section of the population.
A system is non-ergodic when they diverge — typically because the sequence of outcomes can hit an absorbing barrier, a state you cannot exit.
The canonical example:
- 100 people each go to a casino once: if person 28 goes bankrupt, persons 29-100 are unaffected. The ensemble average across the group is informative.
- 1 person goes to the casino 100 times: if they go bankrupt on visit 28, there is no visit 29. The expected value of future rounds is irrelevant — they can't play them.
These two scenarios have the same expected value per round. They have completely different outcomes. The first is ergodic; the second is non-ergodic.
Why it matters:
Most real-world risk — investment, career, business, health — is sequential, not one-time. When the sequence can hit ruin, the expected value calculation (which averages across a population) tells you what happens to the average person, not what happens to you specifically when you hit the bad tail.
Nassim Taleb uses ergodicity to explain why: - Expected value reasoning fails when ruin is possible - The Kelly Criterion (size bets proportionally, never risk ruin) is more rational than expected-value maximization - "Rationality" should mean survival, not utility maximization
The river is four feet deep on average. Knowing the average doesn't tell you whether you survive the crossing.
For the full framework, read Ergodicity Explained or Ergodicity, Ruin, and Rational Risk-Taking.