The Nobel Halo at Thirty-Times Leverage

LTCM's partners included Robert Merton — the very man who generalized Kelly to continuous time, π* = (μ − r)/σ² — and fellow Nobel laureate Scholes. Yet they didn't use Kelly. They used VaR risk control built on a normal-distribution assumption, at 30× leverage — far beyond what the formula advises even with the most optimistic parameters.

In 1998 the fat tail arrived: 90% gone in three months, US$4.6B evaporated, the Fed organized a rescue. Kelly (especially Merton's own 1971 version) carries an implicit warning: when fat tails exist, cut the optimal position sharply. They ignored it.

Every genuinely successful Kelly user — Thorp, Gross, Buffett — uses fractional Kelly to absorb estimation error. F-Star's quarter-Kelly default plus risk caps exist precisely so as not to become the next LTCM.