The Kelly framework, intuitively. Kelly sizing answers a specific question: given a known edge (probability of winning and the size of the win vs the loss), what fraction of capital maximizes the long-run geometric growth rate? The answer for a simple binary bet is f = (p × b - q) / b, where p is the probability of winning, q = 1 - p, and b is the ratio of win-to-loss outcome. For a typical equity setup with 55% hit rate and 1.6x reward / risk, full-Kelly is roughly 23-25% of capital — a number that strikes most practitioners as wildly oversized, and they are right to find it so.
| Sizing approach | Typical fraction | Trade-off |
|---|---|---|
| Full-Kelly | Often 15-30% per high-conviction position | Maximizes growth IF the edge is exactly correct; produces enormous drawdowns when it is not |
| Half-Kelly | Roughly 7-15% per position | Captures most of the long-run growth (estimates suggest ~85%) with about half the drawdown of full-Kelly |
| Quarter-Kelly | Roughly 3-8% per position | Most professional fundamental investors live here — gives up some growth in exchange for survivability and psychological steadiness during drawdowns |
| Fixed-fraction | 1-3% per position regardless of conviction | Sacrifices the ability to express asymmetric edge; survives almost any drawdown sequence; appropriate when edge is highly uncertain |
Why most professionals live between quarter-Kelly and half-Kelly. Three structural reasons. (1) Edge uncertainty: the gap between your estimated edge and your true edge is large in equity markets, and Kelly is unforgiving of over-estimation — a 5-point miss in hit rate can turn a sizing decision from optimal to ruinous. (2) Fat tails: equity returns have more extreme outcomes than the Kelly derivation assumes, and full-Kelly under-weights the cost of those tails. (3) Behavioral durability: a 40% drawdown is mathematically survivable but practically catastrophic — investors capitulate at extremes, forced selling is locked in, and the long-run compound rate suffers more from the behavioral break than from the mathematical setback.
Worked example — quarter-Kelly on a real position. Pelham Holdings setup: bull $80 (35%), base $58 (45%), bear $32 (20%), current price $52. Reward to bull is $28; risk to bear is $20; ratio is 1.4x. Probability-weighted EV: $60.50, implying ~16% expected upside. Full-Kelly suggests ~18% of capital; quarter-Kelly is ~4.5%; half-Kelly is ~9%. A quarter-Kelly 4.5% position is the standard professional size for a single name at this conviction level. If the analyst's calibration curve suggests their high-conviction sizing is overconfident, the right adjustment is to move TOWARD fixed-fraction and AWAY from half-Kelly until the calibration normalizes.
## See also: deeper references - **Risk / reward asymmetry mechanics:** `ptk-4` in `practitioner-toolkit-201` — for the bull / base / bear sizing math that feeds the Kelly fraction. - **Drawdown tolerance and portfolio-level risk:** `risk-1` and `risk-4` in `risk-management-201` — for portfolio-level risk budgeting that constrains single-position sizing. - **Risk-tolerance and utility theory:** `rt-1` and `rt-2` in `risk-management-201` — for the Jensen's-inequality and stochastic-dominance foundations of risk-averse sizing. - **Diversification mathematics:** `port-2` in `portfolio-101` — for why correlated positions cannot be sized independently. - **Loss aversion and the behavioral cost of drawdown:** `bf-2` in `behavioral-finance-201` — for why a mathematically survivable drawdown is often behaviorally fatal.
Sit with the ideas.
Your edge across 80 logged positions shows an average reward / risk asymmetry of 1.6x with a hit rate of 55%. Full-Kelly sizing on a single conviction-level position would suggest about 25% of capital. A peer with the same edge sizes positions at 6% — a quarter-Kelly. Which framing best explains the gap between full-Kelly and the practitioner-grade size your peer uses?