understand / seed
Convexity
Payoff structures where the upside dwarfs the downside
Buying a lottery ticket and lending money to a friend feel like opposite gestures. Underneath they have the same shape, just flipped. The lottery ticket: small reliable cost, occasional large gain. The loan: small reliable gain in interest, occasional large loss when she can’t pay you back. Convex and concave. The two shapes feel similar most of the time and behave catastrophically differently in the tail.
An average can hide the feature that matters most: how gains and losses change at the extremes. Expected value describes repeated trials; exposure also depends on how many trials you can survive.
Some software experiments are convex: a prototype has a capped cost, while a rare success can be reused widely. A large fixed-scope project may have the opposite shape. Picking up nickels in front of a steamroller — Taleb‘s image for a strategy with frequent small gains and rare ruin — is concave.
Convexity is the mathematical heart of antifragility. Antifragile things gain from disorder because their payoff is convex: limited downside, unbounded upside, so volatility is a friend rather than an enemy. The barbell strategy is a way to manufacture convexity when nature didn’t supply it — combine a hard floor with uncapped exposure on a small slice, and you’ve built the shape on purpose.
For any commitment, ask: what happens to my outcome as conditions become more extreme? If the curve bends upward, additional variation helps; if it bends downward, extreme variation hurts. The diagram does not make the decision for you, but it reveals what kind of risk you are taking.
In uncertain domains, small experiments can create convex exposure when their cost is genuinely capped and their discoveries can travel. A grand plan can be concave when many dependencies must all succeed. Bricolage often works this way. So can slack in a schedule: the empty hour has a visible cost and occasionally makes an unplanned opportunity possible.
The dangerous case is an arrangement that looks stable because losses have not appeared yet. Diversification may reduce some risks while leaving shared tail exposure intact. A claimed floor is only real if you can name the mechanism that enforces it. If you cannot draw the downside, you may be looking at a delayed loss rather than a bounded one.