Kelly Criterion

A mathematical framework for bet-sizing when you have a probabilistic edge. The core principle: never risk ruin. Even with a positive expected value, betting too large a fraction of your bankroll can drive you to zero.

---

The Concept

If you have a 51-to-49 edge in a game, the naive response is “bet the whole kitty — I’m statistically favored.” This is wrong. You could still lose and never get to play again.

The Kelly criterion mathematically specifies the optimal fraction of your bankroll to bet on each turn to maximize long-term growth while avoiding ruin. For a 51-to-49 edge, the optimal Kelly bet is much smaller than your total stake.

---

Ergodicity

naval-ravikant connects Kelly to Nassim Taleb’s concept of ergodicity:

What is true for 100 people on average isn’t the same as one person doing the same thing 100 times.

Russian roulette example: Six people each play Russian roulette once. One dies; five win $1B.Expectedvalueperperson:$833M. But one person playing six rounds: certain death, zero dollars. The population average is positive; the individual path is catastrophic.

Applying population-average reasoning to your individual path is a fundamental error. This is why expected value calculations must be adjusted for path-dependence and ruin risk.

---

Reputation as Bankroll

Naval’s most important practical application: ruining your reputation is the same as getting wiped to zero.

The Kelly criterion’s lesson isn’t just about financial bets. In business and life:

• Cutting corners (legal violations, ethical breaches) is a negative-Kelly bet — the upside is small, the downside is total ruin
• Burning bridges is a negative-Kelly bet on your social capital
• Taking on unacceptable personal risk (health, relationships, reputation) for short-term gain violates Kelly

“The number one way people get ruined in modern business is not by betting too much on a venture. It’s by cutting corners and doing unethical or downright illegal things.”

---

Practical Rules

1. Never go to zero. Always preserve optionality; you can’t recover from ruin.
2. Size your bets. Even high-confidence bets should be a fraction of total capital, not everything.
3. Protect your reputation. Treat reputation as a bankroll subject to Kelly constraints.
4. Avoid irreversible decisions. One-way doors are negative-Kelly unless the upside is enormous and the probability of ruin is truly tiny.

---

Connections

• bayes-theorem: Kelly is the bet-sizing companion to Bayesian updating. Bayes tells you the probability; Kelly tells you how much to act on it.
• probability-theory: The formal mathematical underpinning of Kelly requires expected value calculations and knowledge of distributions.
• fallibilism: Kelly criterion is deeply fallibilist — it institutionalizes humility about outcomes. Even when you’re right, you could be wrong enough to ruin everything.
• high-agency: High-agency people pursue goals ambitiously, but Kelly defines the constraint — don’t bet in ways that remove you from the game entirely.
• sunk-cost-fallacy: Kelly and sunk-cost are related — Kelly tells you to size based on future expected value, ignoring what you’ve already committed.
• mental-models: One of Naval’s core microeconomics mental models.
• compound-interest: Kelly maximizes long-term compounding by preventing the catastrophic loss that resets you to zero (from which no compounding can occur).

Sources

• source—how-to-get-rich — Naval’s exposition; Nassim Taleb’s ergodicity framework referenced