Using the Kelly Criterion to Size a Multi-Strategy Portfolio

Summary

This article presents the Kelly criterion as a way to choose leverage and allocate capital among algorithmic trading strategies to maximize long-run compounded growth. Under its simplified single-strategy assumptions, the recommended leverage depends on expected excess return relative to return variance. For multiple strategies, the treatment assumes normally distributed returns with fixed estimates, statistical independence, reinvested profits and returns net of financing and trading costs. It illustrates the calculation with a hypothetical stock and describes how daily rebalancing can require adding exposure after gains and cutting it after losses.

The article stresses that these assumptions are fragile: strategy returns change, can be non-Gaussian or correlated, and parameter estimates are uncertain. It presents full Kelly as an upper bound rather than a direct prescription, noting that excessive leverage can lead to ruin, and identifies half-Kelly as a more conservative practice. Institutional drawdown, liquidity and allocation constraints may also override growth maximization. The worked example demonstrates the mechanics, not evidence that its projected growth will occur in live trading.

Key ideas

• Kelly sizing chooses leverage based on estimated excess returns and volatility to target long-term compounded growth.
• The stated multi-strategy formulation assumes independent strategies with stable, normally distributed returns.
• The approach presumes continuous rebalancing, while actual trading can only approximate that process.
• Parameter uncertainty and non-Gaussian returns can make full Kelly dangerously aggressive.
• Half-Kelly and institutional constraints offer more conservative alternatives to the theoretical allocation.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.