Strategies

Kelly Criterion

A mathematical formula that calculates the optimal percentage of your bankroll to stake on a bet, based on the size of your edge and the odds.

What Is the Kelly Criterion and Why Does It Matter?

The Kelly Criterion is a mathematical formula developed by John L. Kelly Jr. in 1956 that determines the optimal stake size for a bet based on your estimated probability of winning and the odds offered. It aims to maximise the long-run growth of a bankroll while avoiding the risk of ruin.

At its core, the Kelly Criterion answers a deceptively simple question: How much of your bankroll should you risk on a single bet? Stake too little and you'll miss out on profits. Stake too much and you risk catastrophic losses. The Kelly Criterion calculates the mathematically optimal middle ground.

The formula works by finding the bet size that maximises the expected logarithmic growth of your wealth over time. This is why Kelly is sometimes called the "growth-optimal" strategy. Professional bettors, hedge fund managers, and investment firms use Kelly or fractional Kelly variants as a cornerstone of their bankroll management philosophy.

The Core Principle Behind Kelly

The Kelly Criterion is rooted in information theory. Kelly proved mathematically that if you have a genuine edge (your estimated probability is better than the bookmaker's implied probability), repeatedly betting a specific fraction of your bankroll will grow your wealth faster than any other strategy in the long run.

The key insight is this: the larger your edge, the larger your optimal stake. If you have a 60% win probability on odds of 2.0, your edge is substantial, and Kelly recommends a bigger bet. If your edge is only 52%, Kelly recommends a much smaller bet. This relationship between edge size and stake size is what makes Kelly so powerful—it automatically scales your bets to match your confidence level.

Aspect Kelly Criterion Level Staking Fixed Unit Betting
Approach Varies stake based on edge size Same bet every time Fixed units (e.g., £10 per bet)
Growth Rate Mathematically optimal long-term growth Linear, predictable growth Linear, predictable growth
Volatility High (especially full Kelly) Moderate Moderate
Risk Level Can be high if edge estimate is wrong Low to moderate Low to moderate
Best For Experienced bettors with accurate probability estimates Casual bettors, risk-averse players Beginners, bankroll protection
Probability Estimation Required Very high accuracy needed Low accuracy needed No edge calculation needed

Who Invented the Kelly Criterion and Where Did It Come From?

John L. Kelly Jr. and Bell Labs (1956)

The Kelly Criterion was born from an unexpected place: telecommunications research at Bell Labs in New Jersey. John L. Kelly Jr., a researcher in information theory, published a paper in 1956 titled "A New Interpretation of Information Rate." The paper wasn't about gambling or betting—it was about optimal signal transmission and noise reduction in telephone networks.

However, Kelly included a mathematical proof that applied to gambling: if you repeatedly make bets where you have an edge, the optimal strategy is to bet a fraction of your bankroll equal to your edge divided by the odds. This fraction maximises the expected value of the logarithm of your wealth—in other words, it maximises long-term growth.

Kelly's paper was highly technical and rooted in probability theory and information science. It took several years for the gambling and investing world to recognise its potential.

From Telecommunications to Gambling and Investing

The practical application of Kelly Criterion began in the early 1960s, after MIT student Ed Thorp read Kelly's paper and realised it could be applied to blackjack. Thorp developed a card-counting system for blackjack and used Kelly's formula to determine his bet sizes. He famously beat the casinos using this approach, making Kelly Criterion one of the first mathematically-proven winning strategies in gambling.

From blackjack, Kelly spread to sports betting and eventually to professional investing. Warren Buffett, one of the world's most successful investors, has publicly attributed much of his capital allocation strategy to Kelly Criterion principles. He uses fractional Kelly to determine how much of his portfolio to allocate to different investments.

Today, Kelly Criterion (or fractional variants) is used by:

  • Professional sports bettors and syndicates
  • Hedge fund managers and quantitative investors
  • Poker professionals and card counters
  • Financial traders and portfolio managers
  • Sophisticated betting exchanges and trading platforms

How Do You Calculate the Kelly Criterion Formula?

Breaking Down the Formula Step-by-Step

The Kelly Criterion formula is deceptively simple:

Kelly % = (bp - q) / b

Where:

  • b = net decimal odds (decimal odds minus 1)
  • p = your estimated probability of winning (as a decimal, e.g., 0.55 for 55%)
  • q = your estimated probability of losing (1 - p)

Let's break this down with an example:

Variable Meaning Example
Decimal Odds The odds offered by the bookmaker 2.40
b (net odds) Decimal odds minus 1 2.40 - 1 = 1.40
p (win probability) Your estimated probability of winning 0.55 (55%)
q (loss probability) 1 minus your win probability 1 - 0.55 = 0.45 (45%)
Kelly % (1.40 × 0.55 - 0.45) / 1.40 (0.77 - 0.45) / 1.40 = 0.229 or 22.9%

This means you should stake 22.9% of your bankroll on this bet if you use full Kelly.

Practical Worked Example with Real Betting Odds

Let's say you're betting on a football match. Team A is offered at 2.10 (implied probability of 47.6%). However, through your own analysis, you believe Team A has a 55% chance of winning. This is your edge.

Step 1: Calculate net odds (b)

  • Decimal odds = 2.10
  • b = 2.10 - 1 = 1.10

Step 2: Set your probability (p)

  • p = 0.55 (your 55% estimate)

Step 3: Calculate loss probability (q)

  • q = 1 - 0.55 = 0.45

Step 4: Apply the formula

  • Kelly % = (1.10 × 0.55 - 0.45) / 1.10
  • Kelly % = (0.605 - 0.45) / 1.10
  • Kelly % = 0.155 / 1.10
  • Kelly % = 0.141 or 14.1%

Interpretation: Full Kelly suggests staking 14.1% of your bankroll on this bet. If your bankroll is £1,000, that's £141. If you prefer half Kelly (which most bettors do), you'd stake £70.50 instead.

Using Kelly Criterion Calculators

Manual calculation is prone to errors. Fortunately, numerous free Kelly Criterion calculators are available online. Simply input:

  1. The decimal odds
  2. Your estimated win probability (as a percentage)
  3. Your bankroll (optional, for displaying stake amounts)

The calculator instantly returns your Kelly percentage and recommended stake. Many also show half Kelly, quarter Kelly, and other fractional variants.

Important: Always double-check your inputs. A small error in probability estimate can lead to significantly different Kelly percentages.


Why Do Most Bettors Use Fractional Kelly Instead of Full Kelly?

The Variance Problem with Full Kelly

Full Kelly maximises long-term growth mathematically, but it comes with a catch: it requires perfectly accurate probability estimates. In the real world, probability estimates are uncertain. You might think a team has a 55% chance of winning, but the true probability might be 52% or 58%. This estimation error is the enemy of full Kelly.

If you overestimate your edge and use full Kelly, you'll experience massive bankroll swings. You might lose 20-30% of your bankroll in a short period before recovering. Psychologically, this is brutal. Most bettors can't stomach the volatility, so they abandon their strategy at the worst possible time (after a losing streak).

This is where fractional Kelly comes in. Instead of betting the full Kelly percentage, you bet a fraction of it—typically half Kelly or quarter Kelly.

Half Kelly, Quarter Kelly, and Other Fractions Explained

Half Kelly means betting 50% of the Kelly recommendation. If Kelly says 14.1%, you stake 7.05% instead.

Quarter Kelly means betting 25% of the Kelly recommendation. If Kelly says 14.1%, you stake 3.5% instead.

The benefit of fractional Kelly:

  • Reduced volatility: Half Kelly reduces bankroll swings by roughly half
  • Protection against estimation errors: If your probability estimate is slightly off, you're protected
  • Psychological comfort: Smaller losses are easier to tolerate, so you're more likely to stick to your strategy
  • Still growth-optimal: You still grow your bankroll significantly faster than level staking

The trade-off:

  • Slower long-term growth: You sacrifice some of the maximum growth rate for stability

Most professional bettors use half Kelly or quarter Kelly as a standard. This balances the desire for growth with practical risk management.

How to Choose Your Kelly Fraction

Your choice of Kelly fraction depends on three factors:

  1. Confidence in your probability estimates: If you're very confident (e.g., you have a proven track record of accurate estimates), use half Kelly or higher. If you're less confident, use quarter Kelly or lower.

  2. Bankroll size: Larger bankrolls can tolerate higher Kelly fractions. Smaller bankrolls should use lower fractions to avoid ruin.

  3. Risk tolerance: Some bettors are comfortable with volatility; others aren't. Choose the fraction that lets you sleep at night and stick to your strategy.

A common approach is to start with quarter Kelly, track your results over 100+ bets, and gradually increase to half Kelly once you've proven your edge.


What Does a Negative Kelly Result Mean?

Understanding Negative Kelly

When the Kelly formula produces a negative number, it means you should not bet. A negative Kelly indicates that the bookmaker's odds are unfavourable relative to your probability estimate.

For example, if you estimate a 50% win probability on odds of 2.0:

  • b = 2.0 - 1 = 1.0
  • p = 0.50
  • q = 0.50
  • Kelly % = (1.0 × 0.50 - 0.50) / 1.0 = 0 / 1.0 = 0%

A Kelly of 0% means no bet should be placed. The odds don't offer any edge.

If you estimate 48% on the same 2.0 odds:

  • Kelly % = (1.0 × 0.48 - 0.52) / 1.0 = -4%

A negative Kelly means the bookmaker's implied probability is better than your estimate. You're betting against the odds. Kelly tells you to pass.

How to Respond When Kelly Is Zero or Negative

When Kelly returns zero or negative:

  1. Pass on the bet. Don't stake anything.
  2. Reassess your probability estimate. Is your analysis correct? Did you miss something?
  3. Look for value elsewhere. There are always other bets with positive Kelly.

This is one of Kelly's most important functions: it's not just a staking formula, it's a filter for bad bets. By eliminating zero or negative Kelly bets, you avoid the trap of betting on unfavourable odds.


How Does Kelly Criterion Compare to Other Staking Plans?

Kelly vs. Level Staking

Level staking (also called flat betting) means betting the same amount on every bet, regardless of odds or edge. For example, you might bet £50 on every selection.

Aspect Kelly Criterion Level Staking
Stake Size Varies based on edge Fixed amount every bet
Long-Term Growth Mathematically optimal if edge is accurate Linear and predictable
Volatility High (full Kelly) or moderate (fractional Kelly) Stable
Bankroll Impact Can double/halve quickly Steady, slow growth
Probability Estimation Requires accuracy Not needed
Ease of Use Requires calculation or calculator Simple mental math
Best For Experienced bettors with proven edge Casual bettors, bankroll protection
Example (10 bets, £1,000 bankroll) Bets vary from £30 to £150 £50 per bet

The advantage of Kelly: If you have an edge, Kelly grows your bankroll much faster than level staking. Over 100+ bets, the difference is dramatic.

The advantage of level staking: Simplicity, stability, and protection if your edge estimates are wrong.

Kelly vs. Proportional Staking

Some bettors use proportional staking, which is a fixed percentage of bankroll on every bet (e.g., always bet 5%). This is simpler than Kelly but less optimal because it doesn't account for your edge size.

Proportional staking grows your bankroll predictably but not optimally. Kelly staking grows it faster if your edge estimates are accurate.


What Are the Key Limitations and Risks of Kelly Criterion?

The Probability Estimation Problem

Kelly's biggest weakness is also its biggest requirement: accurate probability estimation. If you overestimate your edge, Kelly will tell you to bet too much. If you underestimate, you'll miss profits.

Professional bettors spend years developing models and testing their probability estimates. Even then, they often find their estimates are off by a few percentage points. This is why fractional Kelly is so important—it provides a buffer against estimation errors.

If you're just starting out and your probability estimates are unreliable, Kelly can be dangerous. You might think you have a 55% edge when you really have 48%, leading to catastrophic overbetting.

Risk of Ruin and Volatility

Full Kelly can lead to significant drawdowns. Imagine you have a 55% win probability on 2.0 odds (14.1% Kelly), but you hit a losing streak. You might lose 25% of your bankroll in 20 bets before recovering. For many bettors, this is psychologically unbearable.

Risk of ruin is the mathematical probability of losing your entire bankroll. With full Kelly, this risk is non-zero. With fractional Kelly, it decreases significantly. With level staking, it's almost zero (as long as you have an edge).

This is why professional bettors use fractional Kelly: they accept slightly lower long-term growth in exchange for much lower risk of ruin.

Practical Challenges in Real Betting

Kelly assumes:

  • Odds are fixed (but they change constantly)
  • No commission or vig (but bookmakers charge)
  • Infinite divisibility of stakes (but there are minimum and maximum bet limits)
  • Your probability estimates are accurate (they're not, always)

In real betting:

  • Odds change before you place your bet
  • Bookmakers charge commission (especially on exchanges)
  • You can't always stake the exact Kelly percentage
  • Market inefficiencies and sharp movement make odds unreliable

These practical constraints mean even professional bettors can't implement pure Kelly. They use approximations and fractional variants instead.


Is Kelly Criterion Used by Professional Bettors?

Adoption Among Sharp Bettors

Yes, absolutely. Most professional sports bettors and syndicates use Kelly Criterion or a variant as their staking foundation. It's taught at professional betting schools and discussed extensively in betting communities.

However, almost no professional uses full Kelly. The standard is half Kelly or quarter Kelly, sometimes with additional risk management overlays.

Professional bettors understand that Kelly is not a guarantee—it's a framework. The real edge comes from accurate probability estimation. Kelly just tells you how to size your bets given your edge.

How Professionals Modify Kelly for Real-World Conditions

Professional modifications to Kelly include:

  1. Fractional Kelly: Using half or quarter Kelly as standard
  2. Portfolio approach: Treating your entire betting portfolio as one unit, not individual bets
  3. Risk limits: Capping maximum stake size or maximum loss per day
  4. Adjustment for correlation: Reducing Kelly if bets are correlated (e.g., betting on multiple games in the same league)
  5. Dynamic Kelly: Adjusting Kelly fraction based on recent performance and confidence

These modifications make Kelly practical in the real world while preserving its mathematical benefits.


Can Kelly Criterion Be Used for Investing?

Kelly in Portfolio Management

Yes. The Kelly Criterion is used by hedge funds, portfolio managers, and quantitative investors to determine capital allocation. Instead of betting on sports, they're allocating capital to different stocks, bonds, or derivatives.

The principle is identical: if you have an edge (your expected return is higher than the market's implied return), Kelly tells you how much capital to allocate to that position based on your edge size and the risk.

Warren Buffett has said that his capital allocation strategy is based on Kelly principles. He allocates more capital to investments where he has higher conviction (larger edge) and less capital to those where he's less confident.

Differences Between Kelly for Betting vs. Investing

Aspect Sports Betting Investing
Time Horizon Days to months Years to decades
Liquidity High (can exit quickly) Lower (some positions illiquid)
Diversification Limited (maybe 10-20 bets) Extensive (100+ positions)
Probability Estimation Based on analysis and models Based on financial analysis and market data
Edge Source Information, analysis, market inefficiency Valuation, fundamental analysis, market inefficiency
Risk Management Fractional Kelly, stop losses Diversification, hedging, position sizing
Leverage Limited (bankroll constraint) Can use leverage (margin)

Investing allows more diversification, which reduces Kelly's downside risk. This is why full Kelly is more practical in investing than in betting.


Common Misconceptions About Kelly Criterion

"Kelly Guarantees Profits"

False. Kelly is a framework, not a guarantee. If your probability estimates are wrong, Kelly will lead you astray. Kelly only works if you have a genuine edge.

Many bettors think Kelly is a magic formula that automatically makes them money. It's not. It's a tool for managing risk and sizing bets. The hard part—finding bets where you have an edge—is entirely up to you.

"You Should Always Use Full Kelly"

False. Full Kelly is mathematically optimal in the long run, but it's impractical in the real world. Fractional Kelly (half or quarter) is safer and more professional.

Full Kelly assumes:

  • Perfect probability estimates (unrealistic)
  • No transaction costs (false)
  • Infinite bankroll (false)
  • Ability to tolerate extreme volatility (most bettors can't)

Even Kelly himself acknowledged that fractional Kelly is often better in practice.

"Kelly Works If You're a Good Predictor"

Partially true, but incomplete. Kelly requires not just good predictions, but consistent edge. You need to:

  1. Estimate probabilities better than the bookmaker, on average
  2. Do this repeatedly and consistently
  3. Maintain your edge over time as markets evolve

Many bettors are good predictors for a period but then lose their edge. Kelly doesn't protect you from this. It just amplifies whatever edge you have—good or bad.


Frequently Asked Questions

Q: What is the Kelly Criterion formula?
A: Kelly % = (bp - q) / b, where b = decimal odds - 1, p = your win probability, and q = 1 - p. The result is the percentage of your bankroll to stake.

Q: Why do most bettors use fractional Kelly instead of full Kelly?
A: Fractional Kelly (half or quarter) reduces volatility and protects against probability estimation errors. Full Kelly requires perfect estimates and can cause severe bankroll swings.

Q: What if the Kelly formula gives a negative result?
A: A negative Kelly means the bet has no edge. You should not bet. Zero or negative Kelly is a signal to pass on that selection.

Q: Is Kelly Criterion used by professional bettors?
A: Yes, most professional bettors use Kelly or fractional Kelly as their staking foundation. It's a standard tool in professional betting.

Q: How do you calculate Kelly Criterion with decimal odds?
A: Subtract 1 from decimal odds to get b. Multiply b by your win probability (p), subtract your loss probability (q = 1 - p), then divide by b. For example: (1.40 × 0.55 - 0.45) / 1.40 = 0.107 or 10.7%.

Q: What are the main risks of using Kelly Criterion?
A: Overestimating your edge, high variance in bankroll swings, and the requirement for accurate probability estimates. Full Kelly can cause significant drawdowns if estimates are wrong.

Q: Can Kelly Criterion be used for investing?
A: Yes, Kelly Criterion is used in portfolio management and capital allocation. It helps determine what fraction of capital to allocate to different investments based on expected returns.

Q: Who invented the Kelly Criterion?
A: John L. Kelly Jr., a researcher at Bell Labs, developed it in 1956 in a paper titled "A New Interpretation of Information Rate."

Q: What is the difference between half Kelly and full Kelly?
A: Full Kelly stakes 100% of the formula's recommendation. Half Kelly stakes 50%. Half Kelly reduces volatility by half and is safer for practical betting.

Q: How do you know if you have an edge for Kelly Criterion?
A: You have an edge when your estimated win probability is higher than the bookmaker's implied probability. The larger the gap, the larger your edge.

Frequently Asked Questions

What is the Kelly Criterion formula?

Kelly % = (bp - q) / b, where b = decimal odds - 1 (net odds), p = your estimated probability of winning, q = 1 - p (probability of losing). The result is the fraction of your bankroll to stake.

Why do most bettors use fractional Kelly instead of full Kelly?

Full Kelly maximises long-run growth but requires perfect probability estimates. In practice, estimates are uncertain, so bettors use half Kelly or quarter Kelly to reduce variance and protect against estimation errors.

What if the Kelly formula gives a negative result?

A negative Kelly means the bet has no edge — the probability does not justify the odds. You should not bet on that selection. Kelly = 0 or negative always means no bet.

Is Kelly Criterion used by professional bettors?

Yes. Many professional and semi-professional bettors use Kelly or a fraction of Kelly as their staking foundation. It is also used in financial investing and portfolio management for similar reasons.

How do you calculate Kelly Criterion with decimal odds?

Convert decimal odds to net odds (subtract 1). Multiply by your win probability, subtract your loss probability, then divide by net odds. For example, with 2.40 odds (b=1.40), 55% win probability: (1.40 × 0.55 - 0.45) / 1.40 = 0.107 or 10.7%.

What are the main risks of using Kelly Criterion?

The primary risks are: overestimating your edge (leading to overbetting), high variance in bankroll swings, and the requirement for accurate probability estimates. Full Kelly can cause significant drawdowns if your estimates are wrong.

Can Kelly Criterion be used for investing?

Yes. The Kelly Criterion is used in portfolio management and capital allocation for long-term wealth building. It helps determine what fraction of capital to allocate to different investments based on expected returns and risk.

Who invented the Kelly Criterion?

John L. Kelly Jr., a researcher at Bell Labs, developed the Kelly Criterion in 1956. He published it in a paper titled 'A New Interpretation of Information Rate,' which connected information theory to optimal betting strategies.

What is the difference between half Kelly and full Kelly?

Full Kelly stakes 100% of the formula's recommendation. Half Kelly stakes 50% of the recommendation. Half Kelly reduces volatility and risk of ruin by half, making it a more conservative but safer approach for practical betting.

How do you know if you have an edge for Kelly Criterion?

You have an edge when your estimated probability of winning is higher than the bookmaker's implied probability (derived from the odds offered). The larger the gap between your estimate and the bookmaker's, the larger your edge.

Related terms