Skip to main content
The Risk-of-Ruin Equation

The Kelly Criterion: Optimal Bet Sizing

Pomegra Learn

What Is the Kelly Criterion and Why Should Every Trader Understand It?

The Kelly Criterion is the mathematical answer to one of trading's most fundamental questions: given your edge and your capital, how much should you risk per trade to maximize long-term wealth while minimizing bankruptcy risk? Discovered by J.L. Kelly Jr. in 1956 and popularized by gambling theorist Ed Thorp and later by billionaire traders, the Kelly Criterion explained reveals the precise position size that balances aggressive growth with catastrophic ruin avoidance.

Unlike arbitrary rules (risk 2% per trade, or risk a fixed dollar amount), the Kelly Criterion is derived from information theory and proven mathematically to maximize the logarithm of your wealth over infinite time. Trades sized at the Kelly Criterion grow faster than any other sizing rule, yet remain stable enough to avoid ruin. Trades sized above Kelly incur unnecessary bankruptcy risk; trades sized below Kelly leave money on the table.

Quick definition: The Kelly Criterion is a formula that calculates the optimal fraction of your bankroll to risk on each bet, based on your win probability and your payoff odds, to maximize long-term compound growth while minimizing ruin risk.

Key takeaways​

  • The Kelly Criterion formula is: f* = (bp − q) / b, where f is the optimal fraction to risk, p is win probability, q is loss probability, and b is the ratio of profit to loss per bet
  • Full Kelly sizing maximizes long-term growth rate but produces severe short-term drawdowns; most traders use fractional Kelly (half-Kelly, quarter-Kelly) to balance growth and stability
  • Kelly sizing automatically adapts to your edge: if your edge is small, Kelly recommends smaller bet sizes; if your edge is large, Kelly permits larger bets
  • Using too much leverage (overbetting Kelly) creates unnecessary ruin risk; using too little (underbetting Kelly) sacrifices compound growth unnecessarily
  • Kelly is indifferent to whether you win often (high p) with small payoffs or win rarely (low p) with large payoffs, as long as expectancy is identical

The Kelly Formula Derives from Information Theory​

The Kelly Criterion emerged from work on information bandwidth and gambling, not from trading theory directly. J.L. Kelly proved that if you repeatedly wager a fixed fraction of your bankroll on bets with known odds, the fraction that maximizes the growth rate of your capital is:

f* = (b*p - q) / b

Where:
  f* = optimal fraction of bankroll to wager per bet
  p = probability of winning (e.g., 0.55 for 55% win rate)
  q = probability of losing = 1 - p
  b = ratio of profit per win to loss per loss (e.g., 1.5 if you win $1.50 for every $1 risked)

The math behind Kelly is rooted in maximizing the expected logarithm of wealth. Logarithmic utility reflects a fundamental principle: a loss of $1 from $10 hurts more than a loss of $1 from $100. Kelly sizing accounts for this diminishing marginal utility and prevents ruin.

Worked Example: Calculating Your Kelly Fraction​

Imagine a simple system:

  • Win rate: 60% (p = 0.60)
  • Win size: $150 per $100 risked (b = 1.5)
  • Loss size: $100 per bet (loss per loss = 1.0)

Calculate Kelly:

f* = (1.5 * 0.60 - 0.40) / 1.5
f* = (0.90 - 0.40) / 1.5
f* = 0.50 / 1.5
f* = 0.333...
f* ≈ 33.3%

The Kelly Criterion recommends risking 33.3% of your bankroll on each trade. If your bankroll is $100,000, you should risk $33,300 per trade.

This might sound aggressive (and it is), but the math says that any other percentage will either: (a) grow slower in the long run, or (b) incur unnecessary ruin risk.

Why Kelly Produces Maximum Long-Term Growth​

To understand why Kelly is optimal, consider the logarithmic growth rate:

Growth Rate = p * log(1 + f*b) + q * log(1 - f)

This equation expresses how your bankroll grows on average after each bet. If you wager 33% on a 60/40 coin flip with 1.5:1 odds, your expected balance after one bet is:

After bet: 0.60 * (1 + 0.333*1.5) + 0.40 * (1 - 0.333)
         = 0.60 * 1.50 + 0.40 * 0.667
         = 0.90 + 0.267
         = 1.167

Your bankroll grows by 16.7% on average. If you instead wager only 20%, your growth is lower. If you wager 40%, your growth is higher until you hit a losing streak, at which point you face ruin.

Overbetting and Underbetting Kelly​

The Kelly Criterion is the sharp knife of position sizing: perfect in theory but dangerous in practice if you overshoot.

Overbetting Kelly (wagering more than f*):

  • If Kelly says 30% and you wager 35%, you're overbetting. Your short-term growth is marginally faster, but your ruin probability rises sharply. A single bad streak can wipe you out.
  • The relationship is exponential: overbetting 20% (36% instead of 30%) can increase ruin probability by 10x or more.

Underbetting Kelly (wagering less than f*):

  • If Kelly says 30% and you wager 15%, you're underbetting. Your ruin probability drops dramatically, but your growth rate is much slower. You're leaving money on the table.
  • A trader who underbeats Kelly by 50% might take 20 years to accumulate wealth that full Kelly would accumulate in 5 years, but with lower drawdowns and lower ruin risk.

Most traders use fractional Kelly (typically half-Kelly or quarter-Kelly, meaning 50% or 25% of the calculated Kelly percentage) to balance growth with psychological comfort and protection against model error.

Example: Full Kelly vs. Fractional Kelly Growth Paths​

Suppose you have a $100,000 account and your strategy's Kelly Criterion is 25% (meaning you should risk $25,000 per trade).

Full Kelly (25% per trade):

  • Bet 1 (Win): $100,000 → $125,000
  • Bet 2 (Loss): $125,000 → $93,750
  • Bet 3 (Win): $93,750 → $117,187
  • Bet 4 (Loss): $117,187 → $87,890
  • Average after 4 bets (assuming WLWL sequence): grows exponentially in the long run, but short-term drawdowns are severe (down to $87,890 early on)

Half-Kelly (12.5% per trade):

  • Bet 1 (Win): $100,000 → $112,500
  • Bet 2 (Loss): $112,500 → $106,875
  • Bet 3 (Win): $106,875 → $120,234
  • Bet 4 (Loss): $120,234 → $114,222
  • Average after 4 bets: grows more smoothly, smaller drawdowns, but slower final growth

Over 100 bets, Full Kelly grows faster, but Half-Kelly is more psychologically tolerable and offers more margin for error if your edge calculation is wrong.

Kelly and Different Types of Payoff Structures​

The Kelly Criterion adapts beautifully to different trading scenarios. Here are three examples with the same expected value but different payoff structures:

Scenario 1: 55% win, $100 per win, $100 per loss (b = 1.0) Kelly = (1.0 * 0.55 - 0.45) / 1.0 = 0.10 / 1.0 = 10%

Scenario 2: 65% win, $50 per win, $100 per loss (b = 0.5) Kelly = (0.5 * 0.65 - 0.35) / 0.5 = 0.30 / 0.5 = 6%

Scenario 3: 40% win, $200 per win, $100 per loss (b = 2.0) Kelly = (2.0 * 0.40 - 0.60) / 2.0 = 0.20 / 2.0 = 10%

Scenarios 1 and 3 have identical Kelly percentages (10%) despite different win rates, because they have identical expected values. Scenario 2, despite having a 65% win rate, has a smaller Kelly fraction (6%) because each win is smaller relative to losses. Kelly is indifferent to whether you win often or win big; it only cares about edge.

Real-World Example: A Forex Trader's Kelly Sizing​

A currency trader backtests a system and finds:

  • Win rate: 54%
  • Average winner: $250
  • Average loser: $200
  • Risk per trade (in dollars): $200

The payoff ratio is b = 250 / 200 = 1.25.

Kelly calculation:

f* = (1.25 * 0.54 - 0.46) / 1.25
f* = (0.675 - 0.46) / 1.25
f* = 0.215 / 1.25
f* = 0.172 = 17.2%

Kelly recommends risking 17.2% of bankroll per trade. The trader has a $50,000 account, so they should risk 17.2% × $50,000 = $8,600 per trade.

But the trader is nervous. A $8,600 loss on a single trade is a huge psychological hit. They decide to use half-Kelly: 8.6% of $50,000 = $4,300 per trade.

Over 100 trades, half-Kelly will produce slower growth than full Kelly, but it will:

  • Reduce drawdown severity
  • Keep the trader psychologically comfortable
  • Provide a margin for error if the 54% win rate is overoptimistic

Decision Tree: Calculating and Implementing Kelly​

The Hidden Cost of Ignoring Kelly​

Many traders ignore the Kelly Criterion entirely and use arbitrary sizing rules (risk 2% per trade, or risk fixed dollars). This leaves them exposed to two opposite risks:

  1. Underbetting: If your system's Kelly is 20%, but you only risk 1%, you're underbetting by 95%. You'll take 10x longer to accumulate wealth, and you might run out of time, patience, or capital before compounding really kicks in.

  2. Overbetting: If your system's Kelly is 5%, but you risk 10%, you're overbetting by 100%. Your ruin probability explodes. A sequence of losses that should be survivable becomes catastrophic.

Traders who use fixed-dollar sizing ($500 per trade regardless of bankroll size) drift toward overbetting as their account shrinks (because $500 becomes a larger percentage of a $10,000 account than a $100,000 account). This is a recipe for ruin during drawdowns.

Common Mistakes​

  1. Calculating Kelly but ignoring model error — Your Kelly calculation is only as good as your p and b estimates. If your true win rate is 48% but you calculated 52%, your Kelly fraction is too large. Always apply a safety buffer (use 50–75% of calculated Kelly).

  2. Using full Kelly without experiencing severe drawdowns — Full Kelly produces shorter maximum drawdowns than underbetting, but they are still uncomfortable. If you've never experienced a 20% drawdown, you're not ready for full Kelly. Practice with fractional Kelly first.

  3. Changing bet sizes mid-strategy — Kelly works only if you stick to your calculated fraction consistently. If you increase size when winning (chasing) or decrease size when losing (scaling back), you break the Kelly math and lose its protective properties.

  4. Ignoring correlation of losses — Kelly assumes independence (one loss doesn't influence the next). If your losses cluster (system fails on Mondays, or during market gaps), losses are correlated, and your true ruin risk exceeds Kelly's prediction. Underbet Kelly as compensation.

  5. Forgetting transaction costs and slippage — Your backtested p and b might not include commissions, slippage, or market impact. Real trading is less profitable than backtests, so your true Kelly is lower. Always apply a buffer.

FAQ​

Casinos cannot directly ban Kelly betting, but they can refuse service or restrict stakes if they detect consistent profitable betting. In trading, Kelly betting is entirely legal and embraced by quantitative funds.

What Happens If I Use 2× Kelly?​

Overbetting Kelly by a factor of 2 (wagering twice the recommended fraction) roughly quadruples your ruin probability while only marginally increasing growth rate. It's almost never worth it. If you're tempted to 2× Kelly, you're not ready for your current edge.

Should I Recalculate Kelly as My Bankroll Changes?​

Yes. As your account grows or shrinks, your Kelly fraction remains mathematically the same percentage, but the dollar amount changes. If your Kelly is 20% and your account was $50,000, you risked $10,000 per trade. Now your account is $75,000; Kelly says risk $15,000 per trade. Many traders use fixed-fractional position sizing (risk X% per trade) to automatically adjust.

What If My Edge Is Very Small?​

If Kelly calculates to 1% or less, you're operating on a razor-thin margin. Consider whether your edge is real or an artifact of overfitting. If the edge is real, use fractional Kelly (50% or 25%) to account for model risk, and ensure you have substantial capital to compound.

Can I Combine Kelly Betting with Diversification?​

Yes, and this is where professional funds operate. If you have multiple uncorrelated strategies, calculate Kelly for each separately, then use a portfolio approach to allocate capital across them. The combined ruin risk of the portfolio is lower than any single strategy at full Kelly.

How Do I Choose Between Full Kelly and Fractional Kelly?​

Full Kelly if: you have 500+ verified trades, your edge is stable across regimes, your model is built on strong fundamentals (not curve-fitted), and you can psychologically tolerate 20–40% drawdowns. Fractional Kelly (half or quarter) if: you're learning, your data is limited, or you value consistency over maximum growth.

Summary​

The Kelly Criterion is not just an academic formula; it is the mathematically optimal answer to position sizing. By connecting your win rate and payoff odds to the fraction of capital you should risk, Kelly ensures that you grow wealth as fast as any strategy permits without accepting unnecessary ruin risk. Full Kelly is theoretically perfect but practically severe; fractional Kelly balances growth with stability.

The traders who have built and maintained billion-dollar portfolios understand Kelly intimately. They don't always use full Kelly (most use 50% or 75%), but they calculate their Kelly fraction first, understand the cost of deviating from it, and let that number inform their sizing decisions. The alternative—arbitrary position sizing—is essentially roulette disguised as strategy.

Next​

→ Fractional Kelly for Real-World Trading