A bet that can wipe you out once will, given enough time.
Avoid ruin first.
The same good bet, 50 rounds. You choose the stake. The other bets a tenth.
A bet that pays off 9 times in 10 still ends you, if one loss can take everything.
A small chance, taken again and again, stops being small. Survive first: the returns only count if you are still there.
A bet with a 1-in-20 chance a year of losing everything. After 30 years, about 21 of 100 people are still in. The other 79 lost it all, however well they did before.
Anything that can lose more than you have, like heavy borrowing, carries this kind of bet inside it, however rarely it bites.
Small bets you can afford to lose are how you learn.
Worked out on this page
If a bet can wipe you out with any fixed chance, repeating it enough times makes ruin near certain. That is probability, not opinion.
- A New Interpretation of Information Rate J. L. Kelly Jr., Bell System Technical Journal 35(4), 917-926, 1956
- The ergodicity problem in economics Ole Peters, Nature Physics 15(12), 1216-1221, 2019
Checked 2 Nov 2026 · How we grade
- A Strong proof: Many good studies or long official data, from different teams or countries, point the same way.
- B Good proof, with limits: Solid studies agree on the direction, but the size varies, or the data covers few places or few kinds of people.
- C Mixed proof: Some good studies for, some against, or the evidence covers only part of the line.
- D Thin proof: Mostly one book’s argument, one survey or stories. It may be wise, but it has not been tested well.
- = True by the sums: It follows from arithmetic, so no study can overturn it. We still show what real data says.
Nassim Nicholas Taleb, Fooled by Randomness (2001). In our words, not quoted.
See it in numbers: Leverage (From Zero)
Next · Red Line 15Don’t put everything on one outcome.