one life to live · metaphor 38 of 100
Given a genuinely good bet — a venture, a move, a marriage of talent and moment — how much of yourself do you stake? Survival dominates expectation: the size of the bet matters more than its goodness.
Two founders leave the same lab with the same good idea and the same real edge. The first is the braver, everyone agrees: she stakes everything she has, twice, and by the third attempt there is nothing left to stake — no savings, no credit, no marriage willing to fund another round. The second stakes a fifth of what he has each time, absorbs two failures without ceremony, and a decade later is still at the table, playing with a stack the first can no longer imagine.
We talk about courage as a question of whether — take the leap or don't. But between the timid life and the ruined one runs a quieter variable: how much. In 1956 a physicist at Bell Labs, John Kelly, showed that this variable has a shape — that for any favorable bet there is a fraction of yourself worth risking, below which you waste the odds and above which the odds, however good, destroy you. Drag the stake and watch.
Everything computes live: the curve and rates are exact; the paths are honest simulations sharing one fixed draw of luck (reroll for another); the halve-before-double figure is a diffusion estimate, exact in the small-stake limit.
The curve of boldness
Every round of this game is favorable at any stake: win 60% of the time, and the expected value of each bet is positive whether you risk 1% or 100%. Expectation says bet more, always more. But you don't live an expectation — you live one sequence of outcomes, compounding. And compounding is ruled by an unforgiving asymmetry: lose half, and you need not half back but a full doubling. Multiplied losses cut deeper than equal multiplied wins heal.
The curve on the left is that truth drawn out. The growth rate g(f) — what your bankroll actually does per round, run long enough — rises with the stake, peaks at the Kelly fraction f* = 2p−1 (20% here), then bends down and, near a 39% stake, crosses zero and keeps falling. Past that line, a bet you win more often than you lose still grinds you toward nothing, with certainty, given time. The bet stays good. The size makes the life bad. Survival dominates expectation.
What to try
Press All-in. Every round each player either doubles or dies, and the dice favor them 60–40. Half the field is gone by round two; by round ten the chart is silence. No path was unlucky — every one of them was playing a winning game, and every one of them is dead, because a single loss at full stake is an exit. Ninety brave founders, one arithmetic.
Now press Timid. Nothing dies — and almost nothing happens. The median crawls; a decade of favorable odds is mostly wasted. Timidity is not the safe answer, only the slow one: its losses are invisible because they are counterfactual, the compounding that never occurred. Then slide the odds: a better edge moves the peak and the ruin line rightward together. Boldness should scale with how good the bet really is — and with nothing else.
Half-Kelly wisdom
Set the stake to 20% — the mathematical optimum — and read the middle dial: even at the peak, there is roughly a one-in-three chance your bankroll halves before it ever doubles. Optimal is not comfortable. Now drag down to 10%, half-Kelly, and watch what each number does: growth keeps about three-quarters of its maximum, while the chance of halving collapses to around one in ten.
That trade is possible because the curve is flat near the top and steep past it: the last increment of boldness buys almost no growth and costs enormous turbulence, while the first increment past the peak costs growth and buys nothing at all. This is why seasoned gamblers and fund managers, who could compute the optimum exactly, deliberately bet half of it. They are buying calm — and, as the caveats below argue, insurance against their own self-regard.
Beyond money
The bankroll needn't be money. It is anything that compounds when you win and can be ruined when you lose: health, staked round after round on an ambitious schedule; trust in a marriage, staked on each hard truth told or withheld; credibility, staked on every bold public claim; energy, staked on each commitment accepted. Each obeys the same curve. A researcher who bets a fraction of her credibility on daring hypotheses compounds a reputation; one who bets all of it on each announcement is one bad result from silence.
Seen this way, the Kelly criterion is a shape for courage: boldness bounded by survival. The question to carry out of this page is "if this goes wrong, am I still at the table?" — that is f, and it is the one you control.
The mapping
| Mathematics | Life |
|---|---|
| bankroll W | The capacity to keep playing — savings, health, trust, reputation, energy. |
| fraction f | How much of yourself this one venture gets; the variable hiding inside "courage." |
| edge p | How good the opportunity really is — not how good it feels, or how good you've told people it is. |
| Kelly f* = 2p−1 | Boldness bounded by survival: the most of yourself a bet ever deserves. |
| f beyond the zero | Overbetting — how genuinely favorable odds still ruin the eager, given time. |
| log growth g(f) | Life compounds: what you keep after each round multiplies what you can try next. |
Where the metaphor tears
The instrument hands you the odds on a dial; life does not. And the punishment is asymmetric: because the curve is flat to the left of the peak and steep to the right, overestimating your edge — the founder's occupational disease — quietly walks you past the summit toward the ruin line, while underestimating it costs only a little growth. Kelly assumes self-knowledge that betting-sized egos rarely have. Half-Kelly is, among other things, a hedge against your own opinion of yourself.
The theorem needs bets that are independent, repeatable, and fractionally divisible. Yours are none of these: you cannot stake 20% of yourself on a marriage, a failed venture bruises the odds of the next one, and some doors open once. Where a bet is indivisible and unrepeatable, Kelly's arithmetic no longer binds — though its spirit survives as a question: what, after the worst case, is left to play with?
Kelly maximizes the long-run growth of the bankroll — full stop. It has no opinion about a person who prizes security over expansion, or one glorious unrepeatable attempt over a lifetime of compounding, or who simply isn't playing to get bigger. Those are not arithmetic errors; they are different objectives, outside the theorem's writ. The math tells you how to grow fastest without dying. It cannot tell you that growing fastest is what your one life is for.