Skip to main content
Risk-of-Ruin Math

Kelly Criterion Basics

Pomegra Learn

What Is the Kelly Criterion and How Do You Use It?

The Kelly Criterion is a mathematical formula that calculates the optimal position size—the exact percentage of your account to risk per trade—that maximizes long-term wealth growth while keeping ruin probability at zero. It answers the seemingly simple question "How much should I risk per trade?" with mathematical rigor. Developed by J.L. Kelly Jr. in 1956 for telecommunications, the formula has become foundational in gambling, investing, and trading.

The Kelly Criterion is not merely a suggestion; it's the mathematically proven position size that achieves the highest long-term compound growth for a given edge. Traders who use it (or a conservative fraction of it) outperform those who size positions intuitively, especially over decades of trading. Yet it requires knowing your edge precisely, which is why many traders misapply it or abandon it when volatility increases.

Quick definition: The Kelly Criterion is the formula that calculates the exact percentage of your account to risk per trade to maximize long-term wealth growth while ensuring ruin probability equals zero.

Key takeaways​

  • The Kelly Criterion formula is: f = [p × b - q] / b, where f is the fraction of account to risk, p is win probability, q is loss probability, and b is the reward-to-risk ratio.
  • Kelly Criterion position sizing achieves the highest possible compound growth rate for a given edge and risk tolerance.
  • Full Kelly sizing produces volatility that many traders cannot tolerate; most professionals use 50–75% of Kelly to reduce volatility while keeping ruin risk negligible.
  • The Kelly Criterion requires accurate edge measurement — backtested or simulated data often overstates win rates and reward-to-risk, leading to overconfident position sizing.
  • Kelly sizing is best for strategies with high sample size (50+ trades) where edge estimation is reliable; it performs poorly with small sample sizes.
  • Most prop trading firms enforce fractional Kelly rules (typically 25–50% Kelly) to protect firm capital while allowing traders to scale with success.

Historical Context: From Gambling to Trading​

J.L. Kelly Jr., a telecommunications researcher at Bell Labs, derived his criterion in 1956 to solve a problem: how much should you bet on a biased coin flip to maximize wealth growth? The answer wasn't "bet everything on the winning side" or "bet equally on all bets." It was a precise percentage that balanced growth against the risk of ruin.

Professional gamblers adopted Kelly sizing in the 1960s and 1970s, discovering that the formula was remarkably robust across different betting scenarios. In the 1980s and 1990s, traders and institutional investors adapted it for equities and derivatives. Warren Buffett and other legendary investors have acknowledged using Kelly-adjacent sizing principles—though they rarely discuss the exact formula publicly.

Today, the Kelly Criterion is taught in graduate finance courses, implemented in algorithmic trading systems, and referenced by almost every serious trader and portfolio manager. Yet paradoxically, most retail traders have never heard of it, positioning intuitively instead and often suffering ruin as a result.

The Kelly Criterion Formula​

The fundamental Kelly Criterion formula for trading is:

Kelly Fraction (f) = [Win Probability × (Reward-to-Risk Ratio) - Loss Probability] / (Reward-to-Risk Ratio)

Or equivalently:

f = [p × b - q] / b

where:
p = probability of a winning trade
q = probability of a losing trade (1 - p)
b = reward-to-risk ratio (average win / average loss)

This formula gives you the exact fraction of your account to risk per trade.

Worked Example 1: Balanced Edge​

Assumptions:

  • Win rate: 55% (p = 0.55)
  • Loss rate: 45% (q = 0.45)
  • Average winning trade: $200
  • Average losing trade: $100
  • Reward-to-risk ratio: 2.0 (b = 2.0)

Calculation:

f = [0.55 × 2.0 - 0.45] / 2.0
  = [1.10 - 0.45] / 2.0
  = 0.65 / 2.0
  = 0.325 = 32.5%

The Kelly Criterion suggests risking 32.5% of your account per trade. On a $100,000 account, that's $32,500 per trade. This position size maximizes long-term wealth growth.

However, 32.5% sizing produces enormous volatility. A losing trade cuts your account by nearly one-third. Most traders can't psychologically or logistically handle this. This is why fractional Kelly is standard.

Fractional Kelly: The Practical Adjustment​

Fractional Kelly reduces the Kelly percentage by a factor—typically 25%, 50%, or 75%. A trader using 75% Kelly multiplies the Kelly fraction by 0.75; one using 50% Kelly multiplies by 0.5.

Using the same example at different Kelly fractions:

Full Kelly = 32.5%
75% Kelly = 32.5% × 0.75 = 24.4%
50% Kelly = 32.5% × 0.50 = 16.3%
25% Kelly = 32.5% × 0.25 = 8.1%

A trader using 50% Kelly on the $100,000 account risks $16,300 per trade instead of $32,500. The volatility is lower, compound growth is slower, but ruin probability remains negligible. Most professional traders and prop trading firms enforce 50% Kelly or lower.

Decision tree​

Worked Example 2: Low Win Rate, High Reward-to-Risk​

Assumptions:

  • Win rate: 40% (p = 0.40)
  • Loss rate: 60% (q = 0.60)
  • Average winning trade: $300
  • Average losing trade: $100
  • Reward-to-risk ratio: 3.0 (b = 3.0)

Calculation:

f = [0.40 × 3.0 - 0.60] / 3.0
  = [1.20 - 0.60] / 3.0
  = 0.60 / 3.0
  = 0.20 = 20%

Even with only a 40% win rate, the 3:1 reward-to-risk produces a 20% Kelly fraction. This demonstrates why reward-to-risk ratio matters so much: a strategy wins less often but wins bigger, and the Kelly formula balances this correctly.

At 50% Kelly, this trader would risk 10% per trade. At 25% Kelly, 5% per trade. Even fractional Kelly sizing allows for positions that reflect the strength of the edge.

Worked Example 3: Negative Edge (The Formula Fails)​

Assumptions:

  • Win rate: 48% (p = 0.48)
  • Loss rate: 52% (q = 0.52)
  • Average winning trade: $100
  • Average losing trade: $100
  • Reward-to-risk ratio: 1.0 (b = 1.0)

Calculation:

f = [0.48 × 1.0 - 0.52] / 1.0
  = [0.48 - 0.52] / 1.0
  = -0.04 / 1.0
  = -0.04 = -4%

A negative Kelly fraction means you have a negative edge—you're expected to lose money long-term. The formula correctly tells you not to trade at all. If Kelly returns negative, your strategy is a net loser, and no position size can fix it.

Why Kelly Sizing Produces Zero Ruin Probability​

The Kelly Criterion is designed such that if you follow it exactly and your edge is real, your account value grows with certainty. The formula mathematically ensures:

  1. Expected value of each trade is positive: With Kelly sizing, the expected win exceeds the expected loss.
  2. Exponential wealth growth: Over time, your account compounds upward.
  3. Ruin is impossible: The formula prevents aggressive enough sizing that ruins you.

This is why Kelly is called optimal: it's not just a good position-sizing rule; it's mathematically proven to be the best rule for maximizing long-term wealth while keeping ruin risk at zero.

The Catch: Knowing Your True Edge​

The Kelly Criterion is only as good as your edge estimate. If your backtest shows 55% win rate but live trading shows 48%, your Kelly sizing was too aggressive. This is the primary failure mode of Kelly positioning in practice.

Sources of edge overestimation include:

  • Overfitting: The backtest was optimized to historical data and doesn't generalize.
  • Slippage: Backtests assume entry and exit prices that real trading can't achieve.
  • Survivorship bias: Losing periods disappear from the data.
  • Regime change: Market conditions shift, and your edge weakens.
  • Black swan assumptions: Models assume normal distributions, ignoring tail events.

A common practice is to apply a margin of safety: if backtesting suggests a 60% win rate, assume 55% for Kelly calculation. This conservative adjustment reduces Kelly fraction significantly but increases the chance Kelly sizing remains appropriate in live trading.

Practical Implementation: From Kelly to Position Size Dollars​

Once you have your Kelly fraction, convert it to dollars per trade. If your Kelly is 20% and your account is $50,000, you risk $10,000 per trade. But "risk" means the maximum loss you're willing to take on that trade, not the position size itself.

Example:

Kelly Fraction = 20%
Account = $50,000
Risk per Trade = 50,000 × 0.20 = $10,000

If you're shorting a stock and place a stop-loss $5 above your entry:
Position Size = $10,000 / $5 = 2,000 shares

If you're trading options and each contract has $100 max loss:
Position Size = $10,000 / $100 = 100 contracts

The mechanics change by asset class, but the principle is identical: Kelly tells you the dollar amount to risk, and your entry/stop-loss structure determines the position size.

Kelly Criterion vs. Fixed Fractional Sizing​

The Kelly Criterion is one approach to position sizing; fixed fractional position sizing (risking a constant 2% or 3% of account per trade) is another. They differ subtly but importantly:

  • Kelly Criterion: Sizing depends on your specific edge (win rate, reward-to-risk). Changes when edge changes. Produces optimal growth but requires accurate edge measurement.
  • Fixed fractional: Sizing is a constant percentage regardless of edge. Simpler, more conservative, doesn't require precise edge knowledge.

Most retail traders use fixed fractional sizing (2–3% per trade) because they lack reliable edge data. Professional traders often use Kelly or fractional Kelly once they've validated their edge with 50+ trades. Both methods are valid; Kelly is optimally aggressive, fixed fractional is conservatively simple.

Real-World Examples​

Example 1: Institutional Trend-Following Fund

  • Win rate: 42%
  • Reward-to-risk: 2.5:1
  • Kelly calculation: [0.42 × 2.5 - 0.58] / 2.5 = [1.05 - 0.58] / 2.5 = 0.188 = 18.8%
  • Actual sizing: 50% Kelly = 9.4%

The fund uses 9.4% per position to control volatility while keeping ruin risk near zero.

Example 2: Day Trading Strategy

  • Win rate: 52%
  • Reward-to-risk: 1.2:1
  • Kelly calculation: [0.52 × 1.2 - 0.48] / 1.2 = [0.624 - 0.48] / 1.2 = 0.12 = 12%
  • Actual sizing: 50% Kelly = 6%

Even with a narrow edge, 6% position sizing is justified by the Kelly formula.

Example 3: Options Selling Strategy

  • Win rate: 68%
  • Reward-to-risk: 0.5:1 (sell premium, collect small amounts, occasional large loss)
  • Kelly calculation: [0.68 × 0.5 - 0.32] / 0.5 = [0.34 - 0.32] / 0.5 = 0.04 = 4%
  • Actual sizing: 50% Kelly = 2%

High win rate but low reward-to-risk produces small Kelly fraction. The formula correctly reflects that the frequent wins don't offset the occasional large losses.

Common Mistakes​

Mistake 1: Using Full Kelly Immediately A trader calculates full Kelly (25%) and sizes accordingly on trade one. After one losing streak, the drawdown is severe and emotional decision-making follows. Using fractional Kelly (50–75%) reduces this risk.

Mistake 2: Assuming Backtested Edge Equals Live Edge Backtests show 58% win rate; live trading shows 51%. Kelly sizing based on backtested data is too aggressive. Always validate with live data first, using conservative estimates.

Mistake 3: Ignoring Regime Changes A strategy works for 6 months, then market conditions shift and win rate drops to 45%. Kelly sizing remains at its original level, and ruin risk skyrockets. Recalculate Kelly quarterly.

Mistake 4: Using Kelly for Strategies with <30 Trades of Data With only 20 trades of backtest history, your win-rate estimate has huge standard error. Kelly is aggressive with small sample sizes. Use fixed fractional sizing instead until you have 50+ trades.

Mistake 5: Mixing Kelly with Emotional Position Sizing "I feel great about this trade, so I'll use 25% Kelly. I'm unsure about this one, so I'll use 10% Kelly." Inconsistent sizing destroys the Kelly advantage. Apply Kelly consistently to all trades.

FAQ​

What if my Kelly fraction is negative?​

Your strategy has a negative expected value. Don't trade it. Return to the drawing board and refine your entry/exit rules.

Can I adjust Kelly based on market volatility?​

The Kelly formula assumes volatility is incorporated into your reward-to-risk ratio. If volatility spikes, you might widen your stops (reducing position size to maintain fixed-dollar risk). This is a separate risk-management decision, not a Kelly adjustment.

How often should I recalculate Kelly?​

Recalculate quarterly or after every 25–50 trades. If your edge changes materially (win rate drops by >3%), adjust immediately.

Is Kelly Criterion better than fixed fractional sizing?​

For traders with validated edge, yes—it's optimal. For traders with uncertain edge, fixed fractional (2–3%) is safer because it's conservative by design and doesn't require precise edge measurement.

What if I want to reduce Kelly even further, like 10% Kelly?​

You can, but it's leaving money on the table from a pure wealth-growth perspective. However, many traders do this to reduce volatility further, and that's a legitimate risk-tolerance choice. The Kelly formula tells you the optimal growth rate; how much you deviate is your decision.

Does Kelly work for multi-leg trades (spreads, hedges)?​

Yes, but the calculation becomes more complex. You need to measure the win rate and reward-to-risk for the entire multi-leg position, not each leg separately. This is where quantitative traders use portfolio-level Kelly optimization.

Summary​

The Kelly Criterion is the mathematically optimal formula for position sizing that maximizes long-term wealth growth while keeping ruin probability at zero. The formula depends on your win rate, loss rate, and reward-to-risk ratio—all measurable from historical data. Most traders use fractional Kelly (50–75%) rather than full Kelly to reduce volatility while maintaining the ruin-protection benefits. The critical requirement is accurate edge measurement; a backtest showing 60% win rate that underperforms to 48% in live trading means Kelly sizing was too aggressive. For traders who validate their edge rigorously and recalculate quarterly, Kelly Criterion sizing is the gold standard of position sizing.

Next​

Kelly Fraction and Fractional Kelly