A small chance of ruin is not a small risk — once you are out of the sequence, the expected value you would have collected goes with you.
A time average is not an ensemble average. A game where 1% are ruined per round is fine for a population and fatal for anyone playing it repeatedly, which is the situation you are actually in.
Links
Typed edges. Anything below the line was authored on the other note and derived here.
prerequisite-of →Compounding rewards duration far more than rate — the dominant variable is how long you leave it alone.The most useful cross-book link here. Housel says duration is the dominant variable; ergodicity is the formal reason why, and why any absorbing barrier invalidates the whole calculation.
supports →Combine extreme safety with extreme risk and refuse the middle: floor the downside absolutely, leave the upside open.The barbell exists precisely to make the absorbing state unreachable. Without the ruin argument it looks like conservatism; with it, it is the only coherent structure.