Beyond the Questionnaire: Kelly Sizing and the Barbell

The conservative-moderate-aggressive framework above is what your broker’s onboarding flow asks. It is a useful starting frame and a misleading endpoint. It implies that risk is a personality trait — and once you have located the personality, your portfolio follows from a template. Two practical sharpenings the questionnaire does not give you are worth knowing.

Kelly Sizing The Kelly criterion,⁸⁴ originally derived in 1956 by an information theorist at Bell Labs and pressed into investing service by Ed Thorp,⁸⁵ answers a sharper question than the risk-tolerance questionnaire: given an edge with known probability and payoff, what fraction of your bankroll maximizes the long-run geometric (compound) growth rate of your wealth? The Kelly fraction for a binary bet is

where $b$ is net odds (payoff per dollar risked), $p$ the probability of winning, and $q=1-p$. The same idea generalizes to continuous returns and to portfolios: the optimal weight in a risky asset is roughly $\mu ∕{\sigma }^2$, where $\mu$ is the expected excess return and ${\sigma }^2$ is the variance.

Two implications are useful even if you never compute the formula:

Doubling Kelly is ruin, half-Kelly is most of the upside

The growth-rate curve as a function of position size is asymmetric. Below the Kelly fraction you under-grow. Above the Kelly fraction, the geometric return falls off a cliff — at 2x Kelly, expected geometric return drops to roughly zero; beyond that, you go broke with probability one over a long enough horizon. Most professional investors who use Kelly at all use half-Kelly or smaller, accepting roughly 75% of the maximum growth rate in exchange for far more tolerable drawdowns. The same logic argues against the standard “leverage your conviction” advice retail accounts keep finding on YouTube.

Your edge is almost always smaller than you think

Kelly demands that you estimate both probability and payoff. Empirically, retail investors badly overestimate their win rate on individual stock picks, which means the formula — if they applied it honestly — would tell most of them to size near zero. The questionnaire would tell them “aggressive.” One of these answers is correct.

The Barbell The framework Taleb has pushed for two decades is structurally simpler and harder for questionnaires to capture: a portfolio split into two extreme buckets, almost nothing in the middle.⁸⁶ 85–90% of capital sits in maximally safe and liquid instruments — short-dated Treasuries, T-bills, zero-coupon Treasury STRIPS, money market funds with negligible credit risk — chosen for survival, not yield. The remaining 10–15% goes into high-convexity, capped-downside trades where the maximum loss is bounded by the position size and the upside is unbounded: long-dated out-of-the-money options (section “Options”), early-stage venture, a concentrated position in something where you have an actual informational edge.

The point of the structure is not that it produces higher returns on average; sometimes it does and sometimes it does not. The point is that the loss distribution has a hard floor (the 10–15% you might write off entirely) and an open ceiling (one position you happened to be right about funds the rest). The standard “moderate” portfolio — 60% equities, 40% bonds, all of it exposed to the same correlated drawdown in a stagflationary regime, see the fallacies discussion in section “Five Fallacies of Fixed-Income Markets” — has the opposite shape: a soft floor that can fall further than expected, and a ceiling capped by the diversification itself.

These tools are not replacements for the risk-tolerance frame; they are sharpenings of it. The questionnaire tells you whether you sleep at night. Kelly tells you how large a position your edge actually justifies. The barbell tells you how to structure the whole portfolio so that you survive the regimes the questionnaire’s templates were not designed for.

Underneath both tools is a posture worth naming. Kelly and the barbell do not ask you to predict what happens next; they ask you to map how your portfolio behaves across the things that could happen, and to refuse positions whose downside under any plausible scenario is unrecoverable. Pinpoint forecasts of recessions, rate cuts, and earnings prints are mostly noise — the forecasters who get rich are the ones selling the forecasts, not the ones acting on them. What you can do, and what these tools are organized around, is rank your exposures by fragility: which ones break catastrophically under stress, which ones merely sag, and bias the portfolio toward the second kind. The goal is not to be right about the future. It is to remain solvent through whichever version of the future actually arrives.