Investment Methodology: Practical Applications of the Kelly Criterion and Position Sizing Optimization

In the previous chapters of this investment methodology series, we established frameworks for risk-off cash rules (Chapter 1), macroeconomic regime tracking (Chapter 2), Federal Reserve net liquidity formulas (Chapter 3), value chain bottleneck identification (Chapter 4), and dynamic indicators for dollar-cost averaging (DCA) (Chapter 5). Having structured these systems, the next tactical challenge in capital management is determining position sizing: how much capital to allocate to an individual asset relative to the total portfolio when a quantitative edge is identified.

While many investors possess strong security screening and execution systems, they often compromise their long-term compounding potential by failing to optimize position sizing. This chapter analyzes the mathematical mechanics of the Kelly Criterion, which balances win probability and payoff ratios to determine optimal position sizing, and introduces a modified Kelly rule designed to protect capital in real-world markets.

The Gambler's Dilemma: Capital Preservation vs. Growth Optimization

To understand the core of the Kelly Criterion, consider the classic coin-tossing game. Imagine a coin biased in your favor, with a 60% probability of landing heads and a 40% probability of landing tails. The game offers an even-money payoff: if you win, you receive 1x your bet; if you lose, you forfeit your bet.

An investor immediately recognizes the positive expected value of this game. However, if they commit 100% of their capital to every turn, they face a 40% probability of absolute ruin on the very first toss. Conversely, if they allocate a tiny 1% fraction of their capital to each turn, the risk of ruin is eliminated, but the long-term rate of wealth accumulation will be extremely low, failing to outpace inflation.

To solve this dilemma and maximize the long-term growth rate of capital, we turn to the formula developed by John L. Kelly Jr. at Bell Labs:

$$f^* = \frac{p \cdot b - (1 - p)}{b} = p - \frac{1 - p}{b}$$

Here, $f^*$ represents the optimal fraction of total capital to allocate to a single position, $p$ is the probability of winning (win rate), and $b$ is the net odds received on the wager (payoff ratio).

Applying the biased coin example where $p = 0.6$ and $b = 1$, the formula yields $f^* = 0.2$. This indicates that allocating exactly 20% of total capital to each turn maximizes the geometric growth rate of wealth. Exceeding this 20% threshold—known as overbetting—reduces long-term compounding efficiency and increases the risk of capital loss.

Real-World Adjustments: The Shift to Half-Kelly Sizing

While mathematically elegant, the standard Kelly Criterion cannot be applied directly to financial markets without modification. Unlike casino games with fixed probabilities, financial market win rates and payoff ratios are dynamic and subject to cognitive biases that lead investors to overestimate their edge.

Executing the standard Kelly ratio in volatile equity markets can expose a portfolio to severe drawdowns during unexpected market spasms or consecutive loss streaks.

To mitigate these risks, institutional allocators enforce a modified framework known as the Half-Kelly (or Fractional Kelly) model. Under the Half-Kelly rule, the investor allocates exactly 50% of the optimal fraction calculated by the standard formula ($0.5 \cdot f^*$).

By executing this conservative strategy, the long-term geometric growth rate of wealth is reduced to approximately 75% of the theoretical maximum, while the expected portfolio variance and maximum drawdowns are reduced by over 50%. This asymmetric risk-reward profile provides a vital defensive buffer.

Ultimately, position size optimization is an exercise in humility: recognizing that our information and forecasts are incomplete and establishing systematic boundaries to prevent excessive concentration. Chasing maximum growth at the expense of capital preservation is a self-defeating strategy that exposes a portfolio to catastrophic downside risks. Incorporating a Half-Kelly filter allows investors to protect capital and support long-term compound growth through volatile market cycles.

---

Quantitative Sizing Reference

To automate the calculation of optimal position sizes, investors can construct a simple spreadsheet model using the following inputs:

• Expected Win Rate ($p$): Estimated probability of positive return (e.g., 0.55)
• Average Payoff Ratio ($b$): Expected gain relative to potential loss (e.g., 1.5)
• Optimal Kelly Fraction ($f^*$): = (B1 * B2 - (1 - B1)) / B2
• Half-Kelly Allocation: = [Optimal Kelly Fraction] * 0.5

---

Execution Rules for Position Sizing

1. Acknowledge Information Gaps: Assume that even high-conviction ideas carry at least a 40% probability of error.
2. Enforce Sizing Limits: Commit no more than half of the standard Kelly calculation (the Half-Kelly rule) to any single position. Restrict individual stock weightings to a maximum of 10% of total portfolio capital under all circumstances.
3. Automate Risk Controls: Set predefined stop-loss (e.g., -15%) and take-profit (e.g., +30%) thresholds to establish a consistent payoff ratio ($b \approx 2.0$) when structuring allocations.

---

Deep Dive: Mathematical Proof of Half-Kelly Performance

Let $f$ represent the allocated fraction of capital, and let $f^*$ be the optimal fraction derived from the standard Kelly Criterion. The geometric growth rate $g(f)$ of a portfolio can be approximated using a second-order Taylor expansion:

$$g(f) \approx r + f(R_a - r) - \frac{f^2\sigma^2}{2}$$

where $r$ is the risk-free rate, $R_a$ is the expected arithmetic return of the asset, and $\sigma^2$ is the variance of returns. Differentiating $g(f)$ with respect to $f$ yields the optimal standard Kelly fraction $f^* = \frac{R_a - r}{\sigma^2}$.

Substituting the Half-Kelly allocation ($f = 0.5 f^*$) into the growth equation gives:

$$g(0.5f*) \approx r + 0.5f*(R_a - r) - \frac{0.25(f*)^2\sigma^2}{2} = r + 0.5f*(R_a - r) - 0.25 \cdot \frac{f*(R_a - r)}{2}$$ $$g(0.5f*) = r + 0.375 f*(R_a - r)$$

This shows that the Half-Kelly allocation captures exactly 75% of the maximum theoretical growth rate.

Concurrently, the portfolio variance term ($f^2\sigma^2$) is reduced to $(0.5f^*)^2\sigma^2$, representing a 75% reduction in volatility risk. This substantial volatility reduction translates directly to a 50% reduction in expected maximum drawdown (MDD), providing the statistical justification for the Half-Kelly rule in institutional capital management.