Fractional Kelly Criterion: How Quantitative Bettors Maximize Growth While Eliminating Ruin

The Position Sizing Paradox

You can have the sharpest prediction model in the world, identifying real statistical discrepancies every single week, and still lose your entire bankroll in three months.

How is that possible?

Because in quantitative modeling, what you risk is just as important as what you identify. Most recreational bettors use flat unit sizes or, worse, double down after a losing streak (the catastrophic Martingale trap). Institutional quantitative desks, on the other hand, treat bankroll allocation as an optimization problem governed by the Kelly Criterion.

The Full Kelly Formula

Originally developed by Bell Labs physicist John L. Kelly Jr. in 1956, the formula calculates the mathematically optimal fraction of your bankroll to allocate to maximize the expected logarithm of wealth:

f* = (b × p - q) / b

Where: - f* is the fraction of current bankroll to allocate. - b is the decimal odds minus 1 (the net decimal payout). - p is the estimated true probability of winning from your model. - q is the probability of losing (1 - p).

For example, if your model estimates a team has a 55% chance of winning at even-money odds (+100 or decimal 2.0, where b = 1):

f* = (1 × 0.55 - 0.45) / 1 = 0.10 (or 10% of your bankroll)

The Danger of "Full Kelly"

While full Kelly maximizes long-term logarithmic growth in theory, in practice it is dangerously volatile. Because model probability estimates are subject to parameter uncertainty and estimation errors, overestimating your edge by just a few percentage points under Full Kelly can lead to gut-wrenching 60% to 80% drawdowns.

Why Sharps Use Fractional Kelly

To protect against model estimation variance and severe market swings, quantitative desks almost universally scale down to Fractional Kelly:

- Half-Kelly (0.5× Kelly): Achieves approximately 75% of the theoretical maximum growth rate while slashing portfolio volatility by 50% and dramatically reducing max drawdown risk. - Quarter-Kelly (0.25× Kelly): The institutional gold standard. It captures substantial compound growth while keeping individual position allocations between 0.5% and 2.5% of bankroll, creating a virtually indestructible equity curve.

At EdgeSlate, our members use our built-in position-sizing calibrator to dynamically apply Quarter-Kelly and Half-Kelly constraints across every sport and model board.