Fractional Kelly means using only part of the capital fraction indicated by
full Kelly. If the model produces f*, for example, half Kelly uses
0.5 × f*.
Why it is used
Reducing the fraction lowers exposure and variability relative to full Kelly under the same assumptions. It is a practical response to uncertainty in probabilities, returns, and their stability. No fraction is universally “prudent”: one-half, one-quarter, or another coefficient is a risk choice that requires justification.
Technical distinction
Fractional Kelly is not the same as risking a fixed percentage of the account without estimating a statistical edge. Reducing size also does not impose a guaranteed drawdown limit: if the distribution or estimates are wrong, even a small fraction may be excessive. The theoretical starting point remains the Kelly criterion.
Sources
- Edward O. Thorp, The Kelly Criterion and the Stock Market, 1992 — Covers logarithmic growth, fixed-fraction betting, and application to markets.
- Enzo Busseti, Eungchun Ryu, and Stephen Boyd, Risk-Constrained Kelly Gambling, 2016 — Compares Kelly, fractional Kelly, and explicit constraints on drawdown risk.