PredictPedia Logo

PredictPedia

MarketsWatchlistWiki

Menu

MarketsWatchlistWikiAboutFAQ
Theme
Log InSign Up
PredictPedia LogoPredictPedia

Comprehensive prediction market intelligence powered by AI. Get real-time insights from Polymarket and Kalshi.

Markets

  • Polymarket
  • Kalshi
  • Watchlist
  • Politics
  • Sports

Resources

  • About
  • FAQ
  • Help Center
  • API Docs
  • Blog

Legal

  • Privacy Policy
  • Terms of Service
  • Disclaimer
  • Contact

© 2025 PredictPedia. All rights reserved.

Data from Polymarket & Kalshi

Live Data
  1. Home
  2. →
  3. Wiki
  4. →
  5. Kelly

Kelly Criterion

Definition

The Kelly Criterion is a bankroll sizing rule that maximizes long-run growth when you have a known edge; can be scaled (e.g., half‑Kelly).

The Formula

For a binary bet with win probability p and odds b:1:

Kelly % = (bp - q) / b

Where:

  • p = probability of winning
  • q = probability of losing (1 - p)
  • b = odds received (payout ÷ stake)

Prediction Market Version

For prediction markets where contracts pay $1:

Kelly % = (p - price) / (1 - price)

Where:

  • p = your estimated probability
  • price = current Yes contract price

Example

Market: "Will it rain tomorrow?"

  • Market price: $0.40 (40%)
  • Your estimate: 60% (p = 0.60)
  • Edge: 0.60 - 0.40 = 0.20 (20%)

Kelly calculation:

Kelly % = (0.60 - 0.40) / (1 - 0.40)
Kelly % = 0.20 / 0.60
Kelly % = 0.333 (33.3%)

Action: Bet 33.3% of your bankroll

Why Kelly Works

Maximizes Growth

  • Optimal between too aggressive and too conservative
  • Grows bankroll fastest in long run
  • Mathematically proven optimal

Prevents Ruin

  • Never bets entire bankroll
  • Scales with edge and confidence
  • Built-in risk management

Compound Growth

  • Winners compound
  • Losers don't destroy bankroll
  • Long-term wealth maximization

Kelly Fractions

Full Kelly (1.0x)

  • Most aggressive
  • Maximum growth rate
  • Highest volatility
  • Can feel scary

Half Kelly (0.5x)

  • Most popular
  • 75% of full Kelly growth
  • Much lower volatility
  • Better sleep at night

Quarter Kelly (0.25x)

  • Conservative
  • ~50% of full Kelly growth
  • Very stable
  • Good for beginners

Example Comparison

Scenario: 60% edge, $10,000 bankroll

| Strategy | Bet Size | Growth Rate | Volatility | |----------|----------|-------------|------------| | Full Kelly | $3,330 | 100% | High | | Half Kelly | $1,665 | 75% | Medium | | Quarter Kelly | $833 | 50% | Low | | Fixed $1,000 | $1,000 | Varies | Fixed |

Practical Application

Step 1: Estimate Probability

Your independent analysis:

True probability = 65%

Step 2: Check Market Price

Current Yes price:

Market price = $0.45 (45%)

Step 3: Calculate Edge

Edge = 0.65 - 0.45 = 0.20 (20% edge)

Step 4: Calculate Kelly

Kelly % = (0.65 - 0.45) / (1 - 0.45)
Kelly % = 0.20 / 0.55
Kelly % = 0.364 (36.4%)

Step 5: Apply Fraction

Using Half Kelly:

Half Kelly = 0.364 / 2 = 0.182 (18.2%)

Step 6: Size Position

With $10,000 bankroll:

Bet = $10,000 × 0.182 = $1,820

Common Mistakes

Overestimating Edge

  • Thinking you have 20% edge when really 5%
  • Results in overbetting
  • Solution: Be conservative with estimates

Ignoring Correlation

  • Kelly assumes independent bets
  • Correlated markets need adjustment
  • Solution: Reduce size for correlated positions

Using Full Kelly

  • Too aggressive for most
  • High drawdowns
  • Solution: Use fractional Kelly

Not Updating

  • Edge changes as prices move
  • Need to recalculate
  • Solution: Dynamic position sizing

When Kelly Doesn't Apply

Unknown Probabilities

  • Can't calculate edge accurately
  • Garbage in, garbage out
  • Alternative: Conservative sizing

Few Bets

  • Kelly optimizes long-run
  • Short-term variance high
  • Alternative: Fixed sizing for one-offs

Liquidity Constraints

  • Can't easily exit
  • Position limits
  • Alternative: Reduce to fit constraints

Emotional Tolerance

  • Can't handle volatility
  • Stress affects decisions
  • Alternative: Use smaller fraction

Kelly for Multiple Markets

Portfolio Kelly

When trading multiple markets:

  1. Calculate edge for each
  2. Consider correlations
  3. Allocate fraction of bankroll
  4. Rebalance as edges change

Correlation Adjustment

  • Reduce Kelly fraction for correlated bets
  • Sum of positions should be less than Kelly
  • More diversification = closer to full Kelly

Advanced: Growth vs Risk

Growth Rate Formula

Growth Rate = r × (1 - σ²/2r²)

Where:

  • r = edge × bet fraction
  • σ = standard deviation

Optimal Trade-off

  • Full Kelly maximizes growth
  • Half Kelly = 75% growth, 50% variance
  • Sweet spot for most traders

Historical Performance

Simulations Show

  • Kelly outperforms fixed sizing
  • Half-Kelly beats Full Kelly risk-adjusted
  • Quarter-Kelly stable but slower
  • Over-Kelly leads to ruin eventually

Related Terms

  • Expected Value
  • Edge
  • Position
  • [Bankroll Management]
Back to Wiki