Detaljerad guide kommer snart
Vi arbetar på en omfattande utbildningsguide för Kelly Criterion Calculator. Kom tillbaka snart för steg-för-steg-förklaringar, formler, verkliga exempel och experttips.
The Kelly Criterion is a mathematical formula that determines the optimal fraction of a bankroll or portfolio to allocate to a favorable bet or investment in order to maximize the long-run logarithmic growth of wealth. Developed by John L. Kelly Jr. at Bell Labs in 1956 in the context of information theory, and independently applied to gambling and investing, the Kelly formula produces the highest long-run compound growth rate among all fixed-fraction betting strategies while ensuring that the probability of ruin is theoretically zero (with continuous betting on favorable odds). For a simple binary bet with probability p of winning (and probability q = 1−p of losing), where a win pays odds b (win b dollars for each dollar bet), the optimal Kelly fraction is: f* = (bp − q) / b = p − q/b. This fraction is the expected value of the bet expressed as a fraction of the bankroll. Betting more than f* (overbetting) reduces compound growth and increases ruin risk; betting less than f* (underbetting) is safe but suboptimal from a long-run growth perspective. For continuous investments (stocks, portfolios), the Kelly formula extends to: f* = (μ − r) / σ², where μ is expected return, r is the risk-free rate, and σ² is the return variance. This result connects directly to the Sharpe ratio: f* = Sharpe Ratio / σ, meaning the Kelly optimal leverage is the Sharpe ratio divided by the volatility. For a typical diversified equity portfolio with Sharpe=0.5 and σ=15%, Kelly suggests 0.5/0.15 = 333% allocation — far above 100%, indicating theoretical leverage. In practice, the full Kelly fraction is rarely used because: (1) real-world edge and probability estimates contain significant uncertainty, making the true edge smaller than estimated (fractional Kelly of 25–50% is common); (2) full Kelly produces severe drawdowns — the expected maximum drawdown before doubling is approximately 50% of wealth; (3) most investors have utility functions that penalize large losses more than logarithmic utility implies. Half-Kelly or quarter-Kelly strategies achieve 75% or 50% of Kelly's long-run growth rate while dramatically reducing drawdown and variance of outcomes. The Kelly Criterion is widely used in sports betting, poker (professional players explicitly use it for bankroll management), commodity trading, and quantitative investment management. Legendary investors Warren Buffett and Charlie Munger have described investment approaches consistent with Kelly principles, and the Criterion underlies Edward Thorp's pioneering work in both blackjack card counting and quantitative hedge fund management.
Kelly Fraction (binary): f* = (b×p − q) / b = p − (1−p)/b Kelly Fraction (continuous): f* = (μ − r) / σ² Growth Rate: g = p × ln(1 + b×f*) + (1−p) × ln(1 − f*)
- 1Estimate the probability p of a favorable outcome (from historical data, model, or edge analysis).
- 2Determine the odds b: net winnings per unit staked in case of success.
- 3Verify positive expected value: EV = b×p − (1−p) > 0. If EV ≤ 0, do not bet (f* ≤ 0 means Kelly recommends no bet).
- 4Calculate Kelly fraction: f* = (b×p − (1−p)) / b.
- 5Apply fractional Kelly if desired: typically f = 0.25 × f* to 0.5 × f* for more conservative capital management.
- 6Size the bet: Bet amount = f* × Total Bankroll.
- 7Recalculate after each bet as bankroll size changes — Kelly is a dynamic strategy with bets proportional to current wealth.
Only 10% of bankroll despite 55% win rate — Kelly is conservative
f* = (1.0 × 0.55 − 0.45) / 1.0 = (0.55 − 0.45) = 0.10 = 10%. Bet amount = 10% × $10,000 = $1,000. Expected value per bet = (0.55 × $1,000) − (0.45 × $1,000) = $550 − $450 = $100. This relatively small fraction (10%) despite a seemingly large edge (55% vs. fair 50%) reflects Kelly's conservatism. Betting the full $10,000 (100%) would maximize expected value per bet but would risk ruin — a run of 10 consecutive losses (probability ≈ 0.3%) would wipe out the bankroll. Kelly's 10% fraction makes the long-run growth optimal while making ruin extraordinarily unlikely.
High variance of poker requires many buy-ins even with positive edge
In poker, the Kelly-related risk of ruin formula: P(ruin) = e^(−2×W×B/V), where W=win rate per hand, B=bankroll in big blinds, V=variance per hand. With win rate 55bb/100 = 0.55bb/hand, variance = 100²/100 = 100 bb²/hand, and bankroll B = 30 buy-ins × 100bb = 3,000bb: P(ruin) = e^(−2×0.0055×3000/100) = e^(−0.33) = 72% — terrible! Professional poker players use 30–50 buy-ins as a minimum. At 50 buy-ins (5,000bb): P(ruin) = e^(−0.55) = 58% — still high. Kelly analysis reveals that poker's high variance requires very deep bankrolls even for winning players.
Full Kelly is extremely aggressive; most practitioners use 25–50% Kelly
f* = (μ − r) / σ² = (0.10 − 0.05) / (0.20²) = 0.05 / 0.04 = 1.25 = 125%. Full Kelly recommends 125% equity allocation (borrowing 25% to invest). This is the Merton optimal portfolio fraction under logarithmic utility. In practice, the true Sharpe ratio and σ are uncertain (estimation error), so actual edge is less than estimated. Using a half-Kelly fraction of 62.5% would achieve roughly 75% of the maximum growth rate while dramatically reducing drawdown risk. Professional quant funds typically use 20–50% of theoretical Kelly to account for estimation uncertainty.
Multi-asset Kelly uses matrix operations on the covariance matrix
Multi-asset Kelly: f* = Σ⁻¹ × α (alpha vector), where Σ is the covariance matrix and α is the vector of excess returns. Asset A has Sharpe 1.0 and relatively low vol — Kelly allocates heavily. Asset B has lower Sharpe and higher vol — smaller Kelly allocation. The positive correlation reduces total portfolio allocation vs. uncorrelated case (where f_total = 55% + 25% = 80%). Practical implementation rounds and constrains: long-only fund would allocate approximately 80% A, 20% B at portfolio level after normalization. The multi-asset Kelly maximizes the portfolio's long-run growth rate given the joint distribution of returns.
Sports betting bankroll management for professional bettors
Poker player stake selection and bankroll management
Quantitative hedge fund position sizing models
Venture capital portfolio allocation across investment opportunities
Trading algorithm capital allocation across concurrent strategies
In practice, this edge case requires careful consideration because standard assumptions may not hold. When encountering this scenario in kelly criterion calculator calculations, practitioners should verify boundary conditions, check for division-by-zero risks, and consider whether the model's assumptions remain valid under these extreme conditions.
In practice, this edge case requires careful consideration because standard assumptions may not hold. When encountering this scenario in kelly criterion calculator calculations, practitioners should verify boundary conditions, check for division-by-zero risks, and consider whether the model's assumptions remain valid under these extreme conditions.
In practice, this edge case requires careful consideration because standard assumptions may not hold. When encountering this scenario in kelly criterion calculator calculations, practitioners should verify boundary conditions, check for division-by-zero risks, and consider whether the model's assumptions remain valid under these extreme conditions.
| Kelly Multiple | Growth Rate (% of Kelly) | Drawdown Severity | Typical User |
|---|---|---|---|
| Full Kelly (1.0×) | 100% | Severe (~50% expected max DD) | Theoretical optimal |
| Half Kelly (0.5×) | 75% | Moderate (~25% max DD) | Aggressive quant funds |
| Quarter Kelly (0.25×) | 44% | Mild (~12% max DD) | Conservative practitioners |
| 10th Kelly (0.1×) | 19% | Very low (~5% max DD) | Risk-averse risk mgmt |
| 2× Kelly (overbetting) | 0% | Extreme — trend to ruin | Danger zone — avoid |
| 3× Kelly (overbetting) | Negative | Near-certain eventual ruin | Catastrophic |
Why is full Kelly rarely used in practice?
Full Kelly maximizes long-run logarithmic growth but produces very aggressive position sizing and severe drawdowns. The expected maximum drawdown before wealth doubles is approximately 50% of current wealth — meaning you should expect to lose half your money at some point before doubling. Most investors have utility functions more risk-averse than logarithmic, placing greater penalty on large losses. Additionally, real-world edge estimates are uncertain — if your true edge is half your estimated edge, full Kelly becomes significant overbetting. Half-Kelly achieves approximately 75% of Kelly's growth rate with much lower variance and maximum drawdown. Professional quant funds typically use 20–50% of Kelly.
What is the relationship between Kelly Criterion and the Sharpe ratio?
For a continuous investment with normally distributed returns, the Kelly optimal leverage is: f* = Sharpe Ratio / σ = (μ − r) / σ². This means the Kelly fraction increases with the Sharpe ratio (higher edge justifies more size) and decreases with volatility (higher vol means more variance risk). The maximum long-run growth rate (Kelly growth) equals the risk-free rate plus (Sharpe²)/2 — so Kelly growth is directly proportional to the square of the Sharpe ratio. This relationship is why the Sharpe ratio is the single most important metric for Kelly-optimal investors: a strategy with twice the Sharpe ratio achieves four times the incremental growth rate.
Can Kelly criterion lead to ruin?
Theoretically, the Kelly criterion has zero probability of ruin in the long run for a favorable bet with continuous positive-probability outcomes (the geometric random walk converges to infinity almost surely). However, in practice, ruin risk exists due to: (1) discrete (not infinitely divisible) bets that can exceed the Kelly fraction in relative terms; (2) model error — the true probability and odds may be less favorable than estimated; (3) streak risk — with full Kelly, long losing streaks cause extreme drawdowns even without ruin; (4) leverage — if the Kelly fraction exceeds 100% (as in stock market applications), borrowed funds create finite ruin possibility. Fractional Kelly strategies are preferred in practice to manage all these risks.
How does the Kelly Criterion handle correlated bets?
For a single bet, Kelly is straightforward. For a portfolio of simultaneously active bets or investments, the multi-asset Kelly uses matrix algebra: f* = Σ⁻¹ × α, where Σ is the covariance matrix of returns and α is the vector of expected excess returns. Positively correlated bets reduce total Kelly allocation compared to uncorrelated bets — the portfolio manager should size down when bets are correlated because correlation amplifies drawdown risk. Negatively correlated bets allow larger total allocation. This is the mathematical foundation for why diversification (negative/low correlation among positions) allows more total capital deployment without sacrificing safety.
What is the 'fractional Kelly' strategy?
Fractional Kelly bets a fixed fraction of the full Kelly amount — typically 25% (quarter-Kelly) to 50% (half-Kelly). Half-Kelly achieves approximately 75% of the maximum long-run growth rate while: (1) reducing portfolio variance to 25% of Kelly variance; (2) halving the maximum expected drawdown; (3) providing substantial buffer against estimation error in edge and probability. The relationship is quadratic: betting at fraction c of Kelly achieves c(2−c) fraction of Kelly's maximum growth rate (always less than 100%), so the growth rate penalty of underbetting is much smaller than the variance reduction benefit. This asymmetry explains why fractional Kelly dominates full Kelly for practical risk management.
Who uses the Kelly Criterion professionally?
The Kelly Criterion has been explicitly adopted or influenced several legendary investment careers. Edward Thorp — a mathematics professor who invented card counting in blackjack — rigorously applied Kelly to both gambling (documented in Beat the Dealer) and his hedge fund Princeton/Newport Partners, which he ran for 19 years with exceptional risk-adjusted returns. Warren Buffett's concentrated position sizing philosophy reflects Kelly thinking: make large bets when you have high conviction and genuine edge. Professional sports bettors and poker players use Kelly explicitly for bankroll management. Quantitative hedge funds like Renaissance Technologies implicitly use Kelly-related optimal fraction concepts in their systematic trading strategies.
What is the 'overbetting' danger with Kelly?
Overbetting — betting more than the Kelly fraction — is worse than underbetting from both growth rate and ruin probability perspectives. At exactly 2× Kelly (double Kelly), the long-run growth rate falls to zero — the account value oscillates around a level without growing. Beyond 2× Kelly, the growth rate becomes negative — the account is virtually certain to eventually lose all value. This stark asymmetry (underbetting is merely suboptimal, overbetting can be catastrophic) makes Kelly an important upper bound on position size. In practice, whenever a model suggests a Kelly fraction very close to or exceeding 100%, treat this as a signal to aggressively apply a fractional Kelly factor due to the heightened overbetting risk.
Proffstips
Run a Monte Carlo simulation of 1,000 paths over 100 bets using full Kelly, half Kelly, and quarter Kelly. The distribution of terminal wealth outcomes visually demonstrates both the superior central tendency of Kelly and the catastrophic downside of overbetting — far more compellingly than formulas alone.
Visste du?
John Kelly Jr. published his criterion in a 1956 Bell System Technical Journal paper titled 'A New Interpretation of Information Rate,' framing the result in terms of information theory and communication channel capacity — not gambling. The paper did not mention gambling at all. Ed Thorp recognized its application to blackjack and later to the stock market, effectively translating Kelly from information theory into practical investment management.
Referenser
- ›Kelly, J.L. (1956): A New Interpretation of Information Rate, Bell System Technical Journal
- ›Thorp, E.O. (2006): The Kelly Criterion in Blackjack, Sports Betting, and the Stock Market
- ›MacLean, Thorp & Ziemba: The Kelly Capital Growth Investment Criterion, World Scientific
- ›Investopedia: Kelly Criterion Definition