Kelly-System Position Sizing
Cap each wager so one adverse outcome cannot destroy your capital
- Difficulty
- Moderate
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 98%
Kelly-System Position Sizing treats capital as a resource that must survive repeated uncertain decisions. Instead of betting as much as possible whenever the odds appear favorable, the decision-maker estimates the edge, potential payoff, and downside, then risks only a bounded fraction of available capital. Gross described learning the principle while counting cards: even with a strong statistical advantage, an oversized bet can fail because individual outcomes remain uncertain. He later carried the lesson into portfolio risk management. The practical value is behavioral as well as mathematical: a predetermined sizing rule prevents excitement, recent wins, or confidence from turning a favorable opportunity into a risk of ruin. In markets, conservative or fractional Kelly sizing is often appropriate because estimated probabilities are imperfect and conditions change.
Origin
Extracted from Masters in Business, where Bill Gross connected Ed Thorp's blackjack methods to his later investment risk management.
Core principles
- 01A favorable edge does not eliminate adverse outcomes
- 02Capital survival takes priority over maximizing one wager
- 03Position size should reflect both advantage and uncertainty
- 04A repeatable sizing rule reduces emotional overbetting
How to run it
- 1
Estimate the Edge
Determine why the expected payoff is favorable and assign realistic probabilities to the possible outcomes.
Pro tip Use a range of probability estimates when the edge is uncertain.
Watch out An unmeasured conviction is not a reliable edge.
- 2
Define the Downside
Calculate how much capital could be lost if the unfavorable outcome occurs, including correlated or hidden risks.
Pro tip Stress-test a worse outcome than the base case.
Watch out Ignoring tail losses can make a safe-looking position ruinous.
- 3
Calculate a Capital Fraction
Translate the estimated edge and payoff into a maximum fraction of capital to risk. Use a conservative fraction when inputs are noisy.
Pro tip Fractional Kelly can reduce volatility and estimation risk.
Watch out Full Kelly sizing can be aggressive when probabilities are uncertain.
- 4
Apply the Cap Consistently
Place no more than the permitted amount even when confidence feels unusually high.
Pro tip Automate or document the rule before the decision becomes emotional.
Watch out A good opportunity can still produce a losing outcome.
- 5
Recalculate as Conditions Change
Update the edge, downside, and position size when prices, probabilities, or available capital move.
Pro tip Shrink the position when evidence weakens rather than waiting for certainty.
Watch out Do not increase risk merely to recover a prior loss.
In the wild
Gross began in Las Vegas with about two hundred dollars and used card counting plus disciplined money management. He varied bets with the odds but kept each wager bounded so a bad hand could not wipe out the stake, eventually building enough winnings to help pay for graduate school.
→ The experience supplied a repeatable risk-management discipline that he later applied at PIMCO.
A manager believes a bond is mispriced but recognizes that liquidity could vanish and the issuer could deteriorate. Rather than making it a dominant holding, the manager applies a conservative Kelly fraction based on estimated upside and stressed downside.
→ The portfolio retains meaningful upside without risking ruin if the thesis fails.
Common mistakes
Confusing Edge with Certainty
Even a statistically favorable wager can lose. Position sizing must account for outcome variance.
Using Overconfident Probabilities
Kelly calculations amplify bad assumptions. Reduce the fraction when the edge is estimated imprecisely.
Chasing Losses
Increasing a wager simply to recover capital breaks the sizing discipline and raises the probability of ruin.
Is it for you?
Best for
It is best for investors and bettors making repeated decisions with measurable but uncertain advantages.
Not ideal for
It is not ideal when probabilities, payoffs, or total downside cannot be estimated credibly.
From the transcript
“It basically it worked off what they all the Kelly system, UM, the system where you can't bet more than two of your um your…”
“it became an instrumental part of risk management for me because it made a lot of sense”
“it provided a system of money management”
From the episode
Bill Gross on Institutional Bond Trading (Podcast)