📚 Stock Market Glossary
Clear, beginner-friendly explanations, real-world analogies, and visual formulas for key stock market terminology.
Kelly Criterion
Quantitative & Portfolio📖 Beginner-Friendly Explanation
Core Concept & Meaning
The Kelly Criterion is a foundational mathematical formula developed in 1956 by Bell Labs researcher John L. Kelly Jr. to determine the optimal percentage of a bankroll to allocate to a favorable bet or investment.
By calculating the exact trade-off between win probability (p) and payoff odds (b), it maximizes the expected value of the logarithmic wealth growth over time while mathematically eliminating the risk of ruin.
Why It Matters & Mechanism
- Preventing Over-Betting & Ruin: Even an edge with 80% win probability will eventually bankrupt a trader if position sizes are too aggressive (e.g., betting 50% per trade), due to inevitable consecutive drawdown streaks.
- Solving Under-Betting Stagnation: Betting too conservatively leaves massive compound growth on the table. Kelly identifies the exact mathematical peak of the capital growth curve.
- Practical Application via Fractional Kelly: Because real-world stock probabilities cannot be estimated with 100% precision, elite quants deploy 'Half-Kelly' (0.5x) or 'Quarter-Kelly' (0.25x) to smooth out drawdowns while capturing 75% to 90% of maximum growth.
Practical Investment Tips & Pitfalls
Legendary quants like Ed Thorp and Jim Simons built their risk management engines on Kelly principles. If the formula outputs a negative fraction (f* <= 0), it signifies negative expectancy, commanding a strict 'No-Trade' decision.
⚖️ Key Comparison at a Glance
| Metric | Full Kelly (1.0x) | Half Kelly (0.5x) | All-In Fixed Staking |
|---|---|---|---|
| Position Allocation | 100% of theoretical Kelly fraction | 50% of Kelly fraction | 100% total portfolio allocation |
| Expected Growth Rate | Theoretical maximum (100%) | Approx. 75% of maximum growth | Guaranteed ruin in long-run |
| Drawdown Volatility | Severe (frequent 50% drawdowns) | Smooth & manageable drawdown curve | Fatal terminal drawdown |
| Model Error Sensitivity | Highly fragile to win-rate miscalculation | Robust against real-world estimation noise | Zero risk management |