Kelly Criterion Betting: How to Size Your Stakes (Kelly vs Flat Stakes)

Once you've mastered basic bankroll management, these advanced techniques can optimize your betting for long-term growth.

Understanding Variance

Variance = the swings in your results, even when betting correctly.

Key Concept:

Even with a 60% win rate at 2.00 odds (excellent!):

• Expected: Win 6/10 bets
• Reality: Sequences vary wildly

Possible 10-bet sequences:

• WWWWWLLLWW (7 wins - lucky)
• WLWLWWLWLL (5 wins - unlucky)
• LLLLWWWWWW (6 wins - wild swings)

All have the same long-term edge, but very different short-term results

Standard Deviation in Betting

For break-even betting (50% win rate at 2.00 odds), the typical range of wins around the expected number (about ±2 standard deviations, roughly 95% of outcomes):

Takeaway: Short-term results tell you very little. After 100 bets at even money, anything from 40 to 60 wins is within normal variance.

Kelly Criterion Deep Dive

Kelly Formula:

f = (bp - q) / b

Where:

• f = fraction of bankroll to bet
• b = decimal odds - 1
• p = probability of winning
• q = probability of losing (1 - p)

Example Calculation:

Your analysis:

• Team has 55% chance to win
• Bookmaker odds: 2.00 (50% implied)

Kelly calculation:

• b = 2.00 - 1 = 1
• p = 0.55
• q = 0.45

f = (1 × 0.55 - 0.45) / 1 = 0.10 = 10% of bankroll

The Problem with Full Kelly

10% stakes = extreme variance

Losing 5 bets in a row at 10% each:

• Start: €1,000
• After 5 losses: €590 (-41%)
• Need +69% to recover!

Full Kelly maximises long-term growth in theory, but the drawdowns are brutal

Fractional Kelly: The Smart Approach

Many experienced bettors use fractional Kelly:

Quarter-Kelly (25% of Kelly):

• Kelly says 10% → Bet 2.5%
• Smooth growth, manageable drawdowns
• A sensible choice for most bettors

Half-Kelly (50% of Kelly):

• Kelly says 10% → Bet 5%
• Good growth, moderate variance
• For experienced bettors confident in their edge estimates

Comparison Over 1000 Bets (Theoretical):

Assuming every bet has a true 52.5% win probability at 2.00 odds (a 5% edge, so full Kelly = 5% of bankroll), median outcomes from the standard Kelly maths:

The halving probabilities use the standard continuous approximation for Kelly betting with a known edge. In real betting your edge is uncertain, which makes the aggressive options riskier than this table shows.

Half-Kelly keeps about 75% of full Kelly's growth rate; quarter-Kelly keeps about 44%, with far smaller swings. For most bettors, quarter- to half-Kelly is the best risk/reward.

Calculating Your Edge Accurately

Kelly only works if you can estimate probabilities accurately.

Method 1: Historical Win Rate

Example (hypothetical) - track similar bets:

• Won 57/100 at average odds 1.90
• Implied probability = 52.6% (1/1.90)
• Apparent edge = 4.4 percentage points (57% - 52.6%)

Caution: 100 bets is a small sample. From the variance table above, ±10 wins is normal, so this record is also consistent with no edge at all.

Method 2: Model-Based

Use your own model (for example, built on shots data or xG from a third-party source):

• Model predicts 60% win probability
• Bookmaker odds 1.80 = 55.6% implied
• Edge = 4.4 percentage points (60% - 55.6%)

Problem: Overestimating your probability by even 2-3 percentage points can double your Kelly stake or create an edge that doesn't exist

Solution: Be conservative, use fractional Kelly

Portfolio Approach to Betting

Don't put all bets in one category.

Example allocation (adjust to where your edge is):

1. Bet Types

• Match winners (40%)
• Over/Under (30%)
• Player props (20%)
• Other (10%)

2. Leagues

• Premier League (30%)
• La Liga (20%)
• Bundesliga (20%)
• Serie A (15%)
• Other leagues (15%)

3. Odds Ranges

• Low odds 1.30-1.70 (30%)
• Mid odds 1.70-2.50 (50%)
• High odds 2.50+ (20%)

Why diversify?

• Reduces correlation
• Smooths variance
• If one category performs poorly, others can compensate

Correlation Management

Key insight: Some bets are correlated, increasing risk.

Highly Correlated (Avoid Betting Both):

• Team to win + Over 2.5 goals
• Player to score + Team to win
• Multiple player props on the same team

Example:

Bad portfolio (high correlation):

• Liverpool to win
• Salah to score
• Over 2.5 goals
• Liverpool -1 handicap

All 4 bets are likely to win or lose together = much bigger swings than four independent bets

Better portfolio (low correlation):

• Liverpool to win
• Man City to win
• Atalanta Over 2.5
• Bayern player prop

Independent bets = true diversification

Dynamic Bankroll Adjustment

If you stake a percentage of your current bankroll, your stakes adjust automatically:

After a Winning Run (5+ units profit):

• Stakes rise naturally with the bankroll
• Don't add extra on top - a hot streak is not proof of a bigger edge
• Keep following your process

After a Losing Run (5+ units loss):

• Stakes fall naturally with the bankroll
• Reassess your process (are your edge estimates realistic?)
• Consider reducing volume while you review

Flat Period (±2 units):

• Maintain standard stakes
• Keep tracking
• Stay patient

Important: Changes should be gradual and based on your process, not on short-term results

Bankroll Growth Targets

Set realistic expectations:

Exceptional short-term returns are far more often variance than skill.

Advanced Tracking Metrics

Beyond win rate, track (a spreadsheet works well):

1. ROI by Category

• Which bet types are most profitable?
• Which leagues perform best?
• Which odds ranges work?

2. Closing Line Value (CLV)

• Did the odds move in your favour after you bet?
• Consistently positive CLV = a strong sign of good bets long-term

3. Expected Value (EV)

• What was the expected profit per bet?
• Actual profit vs EV shows variance vs skill

4. Longest Losing Streak

• How many units did it cost?
• Can your bankroll survive 2× that?

Example (Hypothetical): A Year with a Bad Quarter

Starting bankroll: €10,000

Strategy:

• Fractional Kelly staking (around 1-2% of bankroll per bet)
• Diversified portfolio
• Every bet logged in a spreadsheet

Year 1:

Q1: €10,000 → €11,200 (+12%)
Q2: €11,200 → €12,800 (+14%)
Q3: €12,800 → €12,400 (-3% - bad quarter)
Q4: €12,400 → €14,500 (+17%)

End of Year 1: €14,500 (+45%)

The point of the example: even a good year usually contains a losing stretch. Sensible stake sizes let the bankroll survive it.

When to Withdraw Profits

Don't let your bankroll grow indefinitely:

Option 1: Fixed Bankroll

• Keep €10,000 in your betting account
• Withdraw all profits monthly
• Safe, but slower growth

Option 2: 50/50 Split

• Keep 50% of profits in the bankroll
• Withdraw 50%
• Balanced growth + taking profits

Option 3: Full Reinvestment

• Keep all profits in the bankroll
• Compound growth
• Maximum growth, higher risk

Recommendation: Start with Option 2, adjust based on your goals

Frequently Asked Questions (FAQ)

Q: What is Kelly Criterion in betting?
A: Kelly Criterion calculates the bet size that maximises long-term bankroll growth: f = (bp - q) / b, which equals your edge divided by (decimal odds - 1). With a 5% edge at 2.00 odds, Kelly = 5% of bankroll. Many experienced bettors use quarter- or half-Kelly to reduce variance.

Q: Why is Kelly Criterion better than flat stakes?
A: Kelly bets more when your edge is larger and less when it is smaller, and it scales stakes with your bankroll, so it grows the bankroll faster in theory. Flat stakes treat all bets equally. The catch: Kelly is only as good as your probability estimates.

Q: What's the difference between Full Kelly and Fractional Kelly?
A: Full Kelly maximises growth but produces huge swings - in theory there is about a 50% chance of halving your bankroll at some point. Half-Kelly keeps about 75% of the growth rate with much smaller drawdowns; quarter-Kelly keeps about 44% with smaller swings still.

Q: How do I calculate my betting edge?
A: Your estimated win probability minus the bookmaker's implied probability. Example: you estimate 55%, odds are 2.00 (50% implied) = 5 percentage points. Verify over a large sample of bets, and track closing line value, which tells you more than results over small samples.

Q: Is Kelly Criterion too aggressive for beginners?
A: Full Kelly, yes. Start with 1-2% flat stakes until you have evidence of a real edge over a large sample (several hundred bets). Then consider quarter-Kelly. Kelly amplifies both gains AND losses if your edge estimate is wrong.

Master basic bankroll management first →

Bottom Line: Advanced Bankroll Rules

✅ Use quarter- to half-Kelly for the best risk/reward
✅ Be conservative with edge estimates (avoid overconfidence)
✅ Diversify across bet types and leagues
✅ Avoid correlated bets in your portfolio
✅ Track advanced metrics (ROI by category, CLV)
✅ Let percentage staking adjust your stakes - don't chase streaks
✅ Set realistic expectations (a small, real edge over many bets)
✅ Withdraw profits regularly (never let your bankroll become your life savings)

The best analysts combine edge-finding with disciplined bankroll management.

Use StatPair's free Kelly calculator - size your stakes from your probability and the odds, then use StatPair's player and team stats to build better probability estimates.

P.S. Kelly Criterion assumes you know your edge accurately. Even a small overestimate leads to overbetting. Validate your edge over a large sample before using Kelly, and track every bet in a spreadsheet. Try StatPair free for 7 days - stats, AI projections and free Kelly, EV and CLV calculators.