Bankroll-Adjusted Kelly: Advanced Bet Sizing

Standard Kelly Criterion provides optimal bet sizing for a single bet with known edge and odds. But real-world betting involves multiple simultaneous bets, uncertain probability assessments, and varying risk tolerances. Bankroll-adjusted Kelly accounts for these realities through dynamic fraction adjustment, correlation awareness, and psychological risk factors. This expert-level approach optimizes bet sizing for actual betting conditions rather than theoretical perfection.

Risk-Adjusted Kelly Fractions

Adjusting Kelly based on confidence and bankroll state:

Confidence-Based Adjustment:

High Confidence (Very sharp line, clear edge):
- Use 50-75% Kelly (Half to Three-Quarter)
- Strong probabilistic assessment
- Clear information advantage

Moderate Confidence (Good line, likely edge):
- Use 25-50% Kelly (Quarter to Half)
- Reasonable probability estimate
- Typical sharp betting situation

Low Confidence (Weak edge, uncertain):
- Use 10-25% Kelly (Tenth to Quarter)
- Unclear probability
- Marginal opportunities

Bankroll State Adjustment:

Drawdown (Down 25%+ from high):
- Reduce Kelly fraction by 50%
- Preserve capital during variance
- Tighten bet selection

Peak (At or near bankroll high):
- Use standard Kelly fraction
- Normal operations
- Maintain discipline

This dynamic approach prevents over-betting during uncertainty and drawdowns.

Multi-Bet Kelly Allocation

Managing Kelly across simultaneous positions:

Problem: Standard Kelly assumes one bet at a time
Reality: Often have 5-10 bets active simultaneously

Approach 1: Total Kelly Budget
- Calculate Kelly for each bet
- Sum total Kelly allocation
- If total exceeds 50% of bankroll, reduce proportionally

Example:
- Bet A: 10% Kelly ($1,000)
- Bet B: 8% Kelly ($800)
- Bet C: 12% Kelly ($1,200)
- Bet D: 15% Kelly ($1,500)
- Total: 45% Kelly ($4,500)

If using Half Kelly rule, max total allocation is 25% ($2,500)
Reduce each bet: $556, $444, $667, $833

Approach 2: Independent Bet Fractions
- Use Quarter Kelly for each bet when making multiple
- Accounts for correlation risk automatically
- Simpler, more conservative

Approach 3: Priority-Based Allocation
- Rank bets by edge/confidence
- Allocate full Kelly to top bet
- Reduce fractions for subsequent bets
- Ensures capital goes to best opportunities first

Correlation-Adjusted Kelly

Accounting for correlated positions:

Uncorrelated Bets:
- NFL Sunday: Different games, different teams
- Can use standard Kelly on each
- Portfolio variance is sum of individual variances

Correlated Bets:
- Same team in multiple bets
- Same sport/league on same day
- Related outcomes

Adjustment Formula (simplified):
Adjusted Kelly = Standard Kelly × √(1 - correlation)

Example:
Two bets on same team, standard Kelly 10% each
Correlation estimated at 0.5 (moderate)
Adjusted Kelly = 10% × √(1 - 0.5) = 10% × 0.707 = 7.07%

Bet 7% on each instead of 10%, reducing correlated risk.

High correlation (0.75+): Reduce Kelly by 50%
Moderate correlation (0.5): Reduce Kelly by 30%
Low correlation (0.25): Reduce Kelly by 15%

Growth Rate Optimization

Maximizing long-term bankroll growth:

Full Kelly maximizes geometric growth but:
- Requires perfect probability knowledge
- Creates extreme variance
- Psychologically difficult

Growth Rate Comparison (same edge/odds):
Full Kelly: 100% growth rate, 50% drawdown risk
Half Kelly: 75% growth rate, 25% drawdown risk
Quarter Kelly: 50% growth rate, 12% drawdown risk

Optimal fraction depends on:
1. Accuracy of probability assessments
2. Psychological tolerance for variance
3. Time horizon (longer = can handle more variance)
4. Bankroll relative to lifestyle needs

Professional Approach:
- Use Quarter Kelly with high confidence
- Use Tenth Kelly with moderate confidence
- Pass on low confidence bets
- Results in ~40-50% of Full Kelly growth with minimal drawdown risk

This balances growth and sustainability.

Dynamic Bankroll Adjustment

Recalculating Kelly as bankroll changes:

Continuous Adjustment Method:
- Recalculate bankroll after every bet settlement
- Adjust unit size immediately
- Most aggressive, optimal for growth

Periodic Adjustment Method:
- Recalculate weekly or monthly
- Smooth variance in unit sizing
- Psychologically easier

Threshold Adjustment Method (Recommended):
- Recalculate when bankroll changes 20-25%
- Up 25%: Increase units proportionally
- Down 25%: Decrease units proportionally
- Balances responsiveness with stability

Example:
Start: $10,000 bankroll, $100 units (1%)
Grows to $13,000: Increase to $130 units
Grows to $16,000: Increase to $160 units
Drops to $12,000: Reduce to $120 units

This ensures unit size stays proportional to bankroll through growth and drawdowns.

Edge Estimation and Kelly

Dealing with imperfect probability assessment:

Problem: Kelly requires accurate win probability
Reality: All probabilities are estimates

Solution: Confidence intervals

Example:
Odds: +200 (3.0 decimal)
Your probability estimate: 40% (but uncertain)

Confidence Interval: 35% to 45%

Kelly Calculation:
- Lower bound (35%): Kelly = (2 × 0.35 - 0.65) / 2 = 0.025 = 2.5%
- Best estimate (40%): Kelly = (2 × 0.40 - 0.60) / 2 = 0.10 = 10%
- Upper bound (45%): Kelly = (2 × 0.45 - 0.55) / 2 = 0.175 = 17.5%

Conservative Approach: Use lower bound Kelly (2.5%)
Moderate Approach: Use midpoint between lower and estimate (6.25%)
Aggressive Approach: Use best estimate (10%)

This accounts for estimation uncertainty and prevents over-betting when confidence is low.

Advanced Kelly Implementation

Always start more conservative than Kelly suggests - fraction down, never up. Account for all simultaneous positions when calculating total Kelly allocation. Reduce Kelly fractions for correlated bets based on correlation strength. Recalculate bankroll and units at least monthly, more often during high activity. Use confidence intervals for probability estimates to avoid over-betting. Track actual results vs Kelly predictions to calibrate probability assessment skill. Remember that Kelly assumes unlimited bankroll replenishment - real bankrolls can't reload. The goal isn't perfect Kelly implementation - it's using Kelly principles to optimize sizing while managing real-world constraints. Even approximate Kelly betting dramatically outperforms flat betting or arbitrary sizing. Focus on the principles: size to edge, account for variance, adjust for reality.