Home → Research Files → Kelly Criterion Examples

Kelly Criterion Examples for Sports Betting

Kelly Criterion examples are useful only when they connect the formula to real betting decisions: price, probability, bankroll, model error, and the choice to pass.

Last updated: September 5, 2026

Key Takeaways

• Kelly sizing starts after the EV check. If a bet does not clear break-even probability, the correct Kelly output is pass.
• Full Kelly is fragile in sports betting. John Kelly's original growth framework assumes useful probability information; noisy sports probabilities make half Kelly or quarter Kelly more realistic.
• Plus-money bets can produce surprisingly large stakes. That does not mean the edge is more certain; it means the payout ratio is doing more work.
• Favorites need clean probability estimates. Expensive prices leave less room for error once vig and break-even rate are included.
• Caps still matter. A sports betting unit size rule can keep a mathematically positive bet from becoming too large for your bankroll.

The Kelly Formula in Betting Terms

The common two-outcome Kelly formula is:

Kelly's 1956 paper connected betting size to long-run growth when the bettor has favorable information. Edward Thorp later popularized practical Kelly thinking in gambling and markets. Sports betting adds a harder input problem: your probability estimate is rarely known with the precision the formula seems to imply.

That is why this guide treats the Kelly Criterion calculator as a sizing tool, not a pick generator. The order should be: convert the odds, estimate fair probability, check expected value, then use Kelly to decide whether the stake is small, capped, or skipped.

Example 1: A -110 Spread with a 55% Win Estimate

Suppose you have a $1,000 bankroll and believe a -110 point spread wins 55% of the time.

Input	Value	Why it matters
Bankroll	$1,000	The stake is calculated as a percentage of current bankroll.
American odds	-110	Decimal odds are about 1.9091.
Win probability	55%	Your estimate is above the 52.38% break-even rate.
Full Kelly	7.25%	About $72.50 on a $1,000 bankroll.
Half Kelly	3.63%	About $36.30.
Quarter Kelly	1.81%	About $18.10.

The full Kelly number looks aggressive because a three-point edge over break-even is meaningful at -110. The practical question is whether your 55% estimate is truly defendable. If it came from a calibrated model with a long tracked sample, half Kelly might be reasonable. If it came from a subjective lean, a one-unit flat bet or quarter Kelly is more honest.

Example 2: A Plus-Money Underdog at +140

Now assume a bettor has a $2,500 bankroll, finds +140 on an underdog, and estimates the true win probability at 45%.

Metric	Calculation	Result
Decimal odds	+140 converts to 2.40	b = 1.40
Break-even probability	100 / (140 + 100)	41.67%
Full Kelly	((1.40 x 0.45) - 0.55) / 1.40	5.71%
Full Kelly stake	$2,500 x 5.71%	$142.75
Quarter Kelly stake	$142.75 x 25%	$35.69

Plus-money examples are where bettors often get reckless. The payout looks attractive, and the formula rewards the gap between your fair probability and the sportsbook's price. But a 45% estimate on an underdog still needs evidence: injury timing, lineup strength, market movement, matchup fit, and comparable prices across books.

Example 3: A Favorite at -180

Favorites make Kelly feel boring until the probability estimate is very strong. Say a bettor has a $1,500 bankroll, sees -180, and estimates the favorite wins 68% of the time.

Metric	Value	Readout
Break-even probability	64.29%	The price already demands a high hit rate.
Estimated edge	68% - 64.29%	3.71 percentage points.
Full Kelly	10.40%	About $156 on $1,500.
Half Kelly	5.20%	About $78.
Risk check	High price sensitivity	A small probability error can erase the edge.

The full Kelly number can look large because the bettor believes the favorite wins often enough to overcome the expensive price. That is exactly where model error matters. If the true probability is 65% instead of 68%, the bet is barely above break-even. If the fair probability is 63%, it is negative EV.

Example 4: A Player Prop at +105

Player props can be attractive because the market is less efficient than major sides and totals. They can also be dangerous because limits, injury news, role volatility, and correlation can change quickly.

Input	Value	Practical adjustment
Bankroll	$800	Smaller bankrolls need tighter unit caps.
Odds	+105	Decimal odds 2.05; break-even 48.78%.
Estimated probability	52%	Small but positive edge if the estimate is sound.
Full Kelly	6.29%	About $50.30.
Quarter Kelly	1.57%	About $12.58.
Suggested cap	1% to 1.5%	Props deserve extra model-error protection.

This is where the player props research checklist and an AI player prop research workflow help. If the probability estimate depends on stale usage, an uncertain injury report, or a single trend split, the Kelly output should be reduced or ignored.

Example 5: The Correct Kelly Stake Is Zero

Kelly is just as useful when it tells you not to bet. Suppose a bettor estimates a -120 moneyline at 53%.

Check	Value	Decision
Price	-120	Break-even probability is 54.55%.
Your estimate	53%	Below break-even.
Expected value	Negative	The price is not good enough.
Kelly result	Negative fraction	Stake $0.

A negative Kelly number is not a suggestion to bet smaller. It is a pass signal. If you still like the side, the next move is to line shop. A better price can change the decision. The line shopping workflow and vig calculator guide explain why the bettable number matters more than the opinion.

Full Kelly vs Half Kelly vs Quarter Kelly Examples

The fraction you choose should reflect confidence in the probability estimate, not confidence in the team. A bettor with a tested model, clean closing-line history, and large sample can justify more Kelly exposure than someone reading one injury angle before kickoff.

Kelly fraction	Stake from a 6% full Kelly output	Best fit	Main risk
Full Kelly	6.00% of bankroll	Highly calibrated repeatable edges	Severe drawdowns when inputs are wrong
Half Kelly	3.00% of bankroll	Serious bettors with tracked probability work	Still large for thin or correlated edges
Quarter Kelly	1.50% of bankroll	Noisy sports markets, props, emerging models	Slower growth if the edge is genuinely strong
Flat unit cap	Usually 0.5% to 2.0%	Beginners or subjective handicappers	May underbet the rare very strong edge

If you are still building a tracked record, use fractional Kelly as a ceiling. A simple rule works well: calculate quarter Kelly, then cap any single bet at your normal unit size unless you have a documented reason to go higher.

A Practical Kelly Workflow

1. Convert the odds. Use an odds converter so the price and implied probability are clear.
2. Remove market tax when needed. For two-sided markets, use a no-vig odds calculator to understand the market's fair baseline.
3. Estimate your probability independently. Do not copy the sportsbook's implied probability and call it an edge.
4. Run EV first. A positive expected-value betting workflow should come before stake sizing.
5. Calculate full Kelly. Treat it as the theoretical maximum, not the default stake.
6. Cut and cap. Apply half Kelly, quarter Kelly, or a fixed unit cap based on uncertainty.
7. Track the result. Log odds, stake, closing line, model probability, and notes in a bet tracker spreadsheet.

Common Mistakes in Kelly Examples

• Using the formula before confirming EV. Kelly cannot rescue a negative-price bet.
• Confusing confidence with probability. Feeling strong about a side is not the same as estimating it at 55%.
• Ignoring vig. The sportsbook's hold raises the hurdle and can make a small edge disappear.
• Overbetting correlated positions. Five props from the same game script are not five independent Kelly opportunities.
• Keeping bankroll static after losses. Kelly stakes should be based on current bankroll, not the peak number you wish you still had.
• Skipping responsible limits. Bankroll math is not a reason to bet money needed elsewhere.

Sources and Further Reading

• John L. Kelly Jr., A New Interpretation of Information Rate, Bell System Technical Journal, July 1956.
• Edward O. Thorp, The Kelly Criterion in Blackjack, Sports Betting, and the Stock Market.
• National Council on Problem Gambling, problem gambling resources and helpline information.

Bottom Line

Kelly examples are best used as a discipline drill. If the bet is not positive EV, stake zero. If the edge is real but uncertain, use fractional Kelly or a fixed unit cap. If the full Kelly number shocks you, that is not a bug. It is the formula reminding you to inspect the probability estimate before the stake gets dangerous.