Quarter Kelly vs. Half Kelly: Which is Better?

The Full Kelly Criterion offers a mathematical absolute: it provides the maximum possible compound growth rate for a series of bets with a known edge. If you risk more than Full Kelly, you enter the zone of negative asymmetry—you assume more risk for less reward, mathematically guaranteeing eventual ruin. If you risk exactly Full Kelly, you maximize your geometric growth.

So why does virtually every successful quantitative hedge fund on the planet refuse to trade Full Kelly?

Because the cost of maximum theoretical growth is maximum practical volatility. A Full Kelly algorithm operates at the absolute edge of disaster. It is mathematically normal for a Full Kelly system to experience drawdowns exceeding 50%, and in severe variance events, drawdowns of 90%. To survive the reality of live markets, you must sacrifice the peak of the growth curve for stability. This is the realm of Fractional Kelly, where the debate between Quarter Kelly and Half Kelly dominates risk architecture.

The Asymmetry of the Kelly Curve

To understand fractional sizing, you must first understand the shape of the Kelly curve. The relationship between risk and compound growth is not linear; it is a parabola. This parabolic nature creates a massive mathematical advantage for fractional sizing.

When you cut your risk in half, you do not cut your profits in half. The peak of the curve is relatively flat, meaning you can dramatically reduce your exposure while capturing the vast majority of the theoretical growth.

Look closely at the math above. By utilizing Half Kelly, you slash your portfolio volatility and your Risk of Ruin by 50%, yet you still capture 75% of the absolute maximum possible growth rate. This is arguably the closest thing to a "free lunch" in quantitative finance.

The Case for Half Kelly: The Institutional Standard

Half Kelly is the default operating standard for most established quantitative algorithms. It provides an optimal balance between aggressive compounding and variance protection.

If your algorithm has a massive sample size (e.g., 5,000+ executions in out-of-sample live trading), you have a high degree of confidence in your Win Rate and Expected Value. When estimation error is low, you can afford to push closer to the peak of the curve. Half Kelly allows you to rapidly compound your capital without crossing the threshold into mathematical suicide.

The Case for Quarter Kelly: The Retail Holy Grail

Retail traders hate Quarter Kelly because it feels too slow. They see "25% Risk" and assume they are leaving money on the table. This is a fatal miscalculation. Quarter Kelly is mathematically superior for 99% of retail traders due to one unavoidable factor: Estimation Error.

The Kelly formula assumes you know your exact edge. You don't. Your win rate is based on historical data, which is a lagging indicator. If you backtested a 55% win rate, but the live market regime shifts and your true forward edge is only 51%, the "optimal" Kelly fraction drastically shrinks.

If you were trading Full Kelly or Half Kelly on a 55% assumed edge, and the edge decays to 51%, you are suddenly over-betting. You have unknowingly crossed the peak of the Kelly curve into negative asymmetry. You will bleed capital rapidly.

The Brutal Conclusion

Stop lying to yourself about the certainty of your edge. Unless you are running an institutional HFT firm with a million-trade dataset, your inputs are estimations.

Start with Quarter Kelly. Prove your edge in live markets for 1,000 executions. Prove that you can survive the inevitable mathematical losing streak without manually intervening or changing your algorithm's logic. Only once you have verifiable, friction-adjusted live data that proves your edge is stable should you even consider scaling up to Half Kelly.