Decisions require ergodicity to succeed over time

Most decision frameworks evaluate bets by their expected value: average the outcomes across everyone who takes the bet, then decide. That logic works if you can run the experiment in parallel across many people. It fails if you are one person running it sequentially over time.

The difference is ergodicity. A process is ergodic when the group average equals what any single participant would experience over a long run. Most high-stakes personal decisions are not ergodic. The gamble where 99 of 100 people win a million dollars looks great on average, but the one person who loses everything is out of the game permanently. One catastrophic loss erases all future wins.

The technical name for this is the absorbing barrier problem. Any outcome that removes you from the game permanently (bankruptcy, death, irreversible damage) breaks the math that makes expected value reasoning valid. Once you hit that barrier, there is no long run left to average over.

The practical shift: stop asking "what works on average for a group?" and start asking "what keeps me in the game?" Survival is the precondition for all future upside. Avoiding ruin is not conservative risk management; it is the highest-leverage move available.

	Ensemble average	Time average
What it measures	Group performance at a snapshot	One person's performance over a long run
Ruin matters?	No	Yes: one absorbing barrier ends the game
Decision rule	Maximize expected value	Survival first, then value

Source claim: For an individual making repeated decisions, a small probability of ruin guarantees eventual failure, so survival must take priority over maximizing expected value.