Capital Growth Theory Logarithmic Utility Kelly (1956) & Thorp (1969)

Kelly Criterion Multi-Asset Portfolio Lab

Model continuous multi-asset Kelly optimization ($w^* = \Sigma^{-1}(\mu - r_f\mathbf{1})$), evaluate Fractional Kelly risk-return tradeoffs, inspect logarithmic growth curves, and stress-test cumulative drawdown probabilities.

Capital Strategies & Presets:
Kelly Portfolio Controls
50%
50% = Half-Kelly (institutional standard); 100% = Full Kelly (growth maximizer); >100% = Over-betting danger zone.
2.0x
Maximum allowable gross risky asset exposure ($\sum w_i$).
3.5%
Cash yield and hurdle rate for excess returns.
0.0%
Stress test estimation errors in expected asset returns.
1.0x
Scales all asset volatilities to simulate market stress.
Compound Growth Rate $g(w)$
6.82%
75% of Full Kelly (9.10%)
Total Risky Leverage
78%
22% Cash reserve
Sharpe Ratio
0.46
Ann. Vol: 10.2%
Probability of 50% Drawdown
12.5%
vs 50% under Full Kelly
Expected Time to Double
10.2 yrs
Expected time to double capital
📈
Calculating Kelly multi-asset portfolio diagnostics...
Multi-Strategy Comparison: Full Kelly vs. Fractional Kelly vs. Markowitz
Portfolio Strategy Expected Return Volatility Geometric Growth $g(w)$ Sharpe Total Exposure P(Drawdown ≥ 50%) Time to Double
Sensitivity Matrix 1: Kelly Fraction ($f$) vs. Alpha Drift on Growth Rate ($g(w) \%$)
Fraction ($f$) Drift -3% Drift -1% Base Drift +2% Drift +4%
Expected continuously compounded growth rate $g(w) = \mu_p - \frac{1}{2}\sigma_p^2$. Notice how higher fractions suffer most under negative drift shocks.
Sensitivity Matrix 2: Volatility Shock vs. Risk-Free Rate on Total Leverage (%)
Vol Shock rf 1.5% rf 2.5% rf 3.5% rf 4.5% rf 5.5%
Total risky capital exposure ($\sum w_i$). Rising volatility and higher hurdle rates automatically compress optimal Kelly leverage.

Institutional Quantitative Foundations: The Kelly Criterion & Capital Growth Theory

1. Logarithmic Utility & Capital Growth Maximization

John L. Kelly Jr. (1956) proved that maximizing the expected logarithm of terminal wealth $\mathbb{E}[\ln(W_T)]$ asymptotically outperforms all other betting strategies almost surely:

$$\lim_{T \to \infty} \frac{1}{T} \ln\left(\frac{W_T^{\text{Kelly}}}{W_T^{\text{other}}}\right) > 0 \quad \text{a.s.}$$

In continuous time under geometric Brownian motion, maximizing $\mathbb{E}[\ln(W_t)]$ is equivalent to maximizing the compound growth rate: $g(w) = r_f + w^T(\mu - r_f\mathbf{1}) - \frac{1}{2}w^T\Sigma w$.

2. Continuous Multi-Asset Kelly Vector

Taking the vector gradient $\nabla_w g(w) = (\mu - r_f\mathbf{1}) - \Sigma w = 0$ yields the unconstrained multivariate Kelly portfolio:

$$w^* = \Sigma^{-1} (\mu - r_f \mathbf{1})$$

This endogenously determines both the relative asset weights (identical to the Markowitz Tangency direction) and the optimal total leverage ratio, balancing excess drift against the asset covariance matrix $\Sigma$.

3. The Half-Kelly Principle & Volatility Reduction

Because the growth rate function $g(f)$ is quadratic in the leverage fraction $f$, scaling optimal weights by $f = 0.5$ (Half-Kelly) delivers an asymmetric benefit:

$$g(0.5) - r_f = 0.75 \times (g_{\max} - r_f), \quad \sigma(0.5) = 0.50 \times \sigma_{\max}$$

Half-Kelly captures 75% of maximum excess compound growth while eliminating 50% of portfolio volatility and 75% of variance, drastically buffering against parameter estimation error.

4. Drawdown Dynamics & The Ruin Zone

The probability of suffering a peak-to-trough drawdown of fraction $D$ before doubling wealth scales exponentially with the inverse Kelly fraction:

$$P(\text{Drawdown} \ge D) \approx (1 - D)^{\frac{2}{f} - 1}$$

Full Kelly ($f=1$) incurs a 50% chance of a 50% drawdown. If an investor over-bets ($f > 2.0$), the compound growth rate becomes negative ($g(f) < r_f$), leading to asymptotic ruin despite positive expected arithmetic return.

Kelly Criterion Portfolio Mastery Quiz

1. What mathematical objective does the Kelly Criterion maximize?

2. In continuous multi-asset markets, what is the unconstrained Kelly weight vector?

3. What is the primary institutional advantage of Half-Kelly ($f = 0.50$) over Full Kelly?

4. What happens if an investor over-bets with leverage exceeding twice the Kelly fraction ($f > 2.0$)?

Frequently Asked Questions: Kelly Criterion Multi-Asset Optimization

The Kelly Criterion (John L. Kelly Jr., 1956; Edward Thorp, 1969) is a mathematical framework that determines optimal asset allocation and leverage to maximize the long-term expected logarithmic growth rate of capital $\mathbb{E}[\ln(W_T)]$. In continuous-time multi-asset markets under geometric Brownian motion, the unconstrained optimal Kelly allocation vector is given by $w^* = \Sigma^{-1} (\mu - r_f \mathbf{1})$, balancing expected excess return against portfolio covariance.

Fractional Kelly scales the optimal Kelly allocation by a fraction $f \in (0, 1]$, such as Half-Kelly ($f = 0.50$). Institutional quantitative funds use Half-Kelly because the growth curve is parabolic: Half-Kelly captures 75% of maximum compound growth rate while slashing portfolio volatility by 50% and reducing variance by 75%, providing a massive safety buffer against parameter estimation error and severe tail drawdowns.

While classical Markowitz mean-variance optimization maximizes a single-period quadratic utility function under an arbitrary investor risk aversion coefficient $\gamma$ and fixed 100% unleveraged budget, the Kelly Criterion maximizes multi-period terminal wealth compounding. In continuous time, Kelly corresponds to logarithmic utility ($\gamma = 1$) and endogenously solves for the exact optimal leverage and cash reserve ratio.

Over-betting beyond Full Kelly ($f > 1.0$) increases portfolio volatility while actively decreasing the expected geometric compound growth rate. If leverage reaches twice the Kelly fraction ($f \ge 2.0$), the expected compound growth rate falls below the risk-free rate, and for $f > 2.0$, the compound growth rate becomes negative, guaranteeing asymptotic ruin almost surely despite having a positive expected arithmetic return.

For an asset compounding under a Kelly fraction $f$, the theoretical probability of experiencing a peak-to-trough drawdown of fraction $D$ before doubling wealth is approximately $P(\text{Drawdown} \ge D) = (1 - D)^{2/f - 1}$. Under Full Kelly ($f = 1.0$), there is a 50% probability of suffering a 50% drawdown. Under Half-Kelly ($f = 0.50$), the exponent increases to 3, reducing the 50% drawdown probability to $(0.5)^3 = 12.5\%$.
Related Quantitative Portfolio Simulations
Ledoit-Wolf Shrinkage Lab

Optimal linear covariance shrinkage and Marcenko-Pastur spectral noise bounds.

Resampled Efficient Frontier Lab

Model Michaud Monte Carlo resampled parameters and rank-averaged frontiers under estimation uncertainty.

Maximum Diversification Lab

Optimize Choueifaty diversification ratio (DR) and test equi-correlation properties.

Hierarchical Risk Parity Lab

Machine learning asset allocation using tree clustering, quasi-diagonalization, and recursive bisection.

Risk Parity & ERC Lab

Allocate capital such that every asset contributes equally to total portfolio risk variance.

Fundamental Law of Active Management Lab

Model Grinold & Kahn Information Ratio ($\text{IR} \approx \text{IC}\sqrt{BR}$), transfer coefficient, and breadth compression.