Home

Bet Sizing Principles: Kelly, Flat, and Percentage Systems

Quick note: This guide is for learning. It is not betting or financial advice. Only bet what you can afford to lose. Check the legal age and rules where you live. If betting stops being fun, get help at BeGambleAware, GamCare, or the National Council on Problem Gambling.

Bet sizing means “how much do I stake on each bet?” It matters a lot. Good sizing can protect your bankroll and help steady growth. Bad sizing can wipe you out even if your picks have an edge. In this guide, we explain three common systems: Kelly, flat betting, and percentage betting. We use simple words, clear steps, and real numbers. You will learn when each system helps, where it fails, and how to choose one that fits you.

TL;DR: Quick Comparison

Kelly Criterion: A math rule that tells you the best stake for long-term growth if your edge (your true win chance) is right. It can be very swingy. Many people use fractional Kelly (like half-Kelly) to cut risk.
Flat Betting: Same stake every time (like 1 unit). It is simple and calm but does not scale with your bankroll or your edge.
Percentage Betting: Bet a fixed percent of your current bankroll (like 1–2%). It adjusts after wins and losses. It is smoother than full Kelly but slower than true Kelly when your edge is real.

What Is Bet Sizing and Why It Matters

Your bankroll is the money you set aside for betting. Your edge is your true chance to win minus what the odds imply. Your variance is how much results jump up and down. Your risk of ruin is your chance to go broke. These ideas are key for smart staking.

If you bet too big, one bad streak can kill your roll. If you bet too small, you may waste a real edge. Good sizing balances growth and safety. It also helps your mind. A clear plan can stop tilt and chasing losses.

Tip: Learn basics like expected value (EV), variance, and risk of ruin before you size up.

The Kelly Criterion

The Formula and the Idea

The Kelly rule tells you what fraction of your bankroll to bet on a single wager when you have an edge. In its simple form for fixed odds, it is:

f = (b × p − q) / b

f = fraction of your bankroll to stake
b = decimal odds minus 1 (for 2.50 odds, b = 1.50)
p = your true win chance
q = 1 − p (your true lose chance)

Kelly says: bet more when your edge is big and the payout is good. Bet zero when you have no edge. If your edge is small or your odds are low, bet a small slice.

For background, see the original paper by J. L. Kelly Jr. (1956) A New Interpretation of Information Rate, and the work of Prof. Edward O. Thorp, who used Kelly in practice: Edward O. Thorp. A deep review appears in MacLean–Thorp–Ziemba (SSRN). A clear overview is also on Wikipedia: Kelly Criterion.

When Kelly Works Well

You have a model or method that gives a solid win chance (p).
You bet many times over a long period.
You can handle swings and stick to the plan.
Your bets are not strongly linked (not all on the same team or outcome risk).

Common Problems and Simple Fixes

Edge error: If your p is wrong, Kelly can overbet. Fix: use fractional Kelly (like 1/2, 1/3, or 1/4 Kelly). This cuts drawdowns.
High swings: Full Kelly can be very bumpy. Fix: smaller fraction, caps per bet, and avoid correlated picks.
Market issues: Limits and line moves can block the exact stake. Fix: use the nearest stake below your target; do not round up.
Small samples: Kelly needs many trials. Fix: log results, review often, and be humble with your edge early on.

Worked Kelly Example

Say odds are 2.10 (decimal). Then b = 1.10. You think the true win chance is p = 0.52 (52%). So q = 0.48.

f = (1.10 × 0.52 − 0.48) / 1.10 = (0.572 − 0.48) / 1.10 = 0.092 / 1.10 ≈ 0.0836.

So full Kelly says stake about 8.4% of your bankroll. If your roll is $1,000, the stake is about $84. Half-Kelly is about $42. If your p was too high, full Kelly could hurt. Half-Kelly gives a safer ride.

Kelly Pros and Cons

Pros: Best growth in theory; adapts to both odds and edge; sets stake to zero if no edge.
Cons: Needs accurate p; very swingy; hard to use with limits; easy to misuse with small data.

Flat Betting

Flat betting means you stake the same amount on every pick. For example, you pick a unit of $10 and always bet 1 unit.

Example: Bankroll $1,000. Stake $10 per bet, no matter the odds. If you have a small edge, you will grow slowly. If you have no edge, you will likely lose slowly. But your swings will be small, and your plan is easy to follow.

Pros: Very simple; helps discipline; low stress. Cons: Does not scale with your bankroll or your edge; may waste strong spots; may bet too much on weak spots by mistake.

Flat betting can work for beginners and for testing a new method. It keeps things steady while you learn. You can later move to a percent plan or to fractional Kelly if your estimates get better.

Percentage (Proportional) Betting

Percentage betting means you stake a fixed percent of your current bankroll each time. Many people use 1% to 2%. This way, your stake goes down after losses and goes up after wins.

Example: Bankroll $1,000. Stake 2% = $20 on the first bet. If you lose, bankroll is $980, next stake is 2% of $980 = $19.60. If you then win at even odds, bankroll goes to $999.60, next stake is 2% of that ≈ $19.99. The plan adjusts on its own. This protects you in bad runs and lets you scale in good runs.

Pros: Simple rules; adapts to your roll; smaller swings than full Kelly. Cons: Not “optimal” like true Kelly when the edge is known; still needs an edge to grow; can be slow if percent is very small.

Many long-term bettors like 0.5% to 2% per bet, based on risk comfort, sport, and sample size. A cautious start is wise, then adjust with data.

Side-by-Side Comparison

Edge needed?	Yes, and it must be accurate	Helps with discipline; edge still needed to profit	Edge helps; adapts even if edge is uncertain
Volatility	High (full); moderate (fractional)	Low	Moderate
Growth potential	Highest in theory (full Kelly)	Low to moderate	Moderate
Ease to use	Harder (needs p and math)	Easiest	Easy
Good for	Model users; long horizon; stable edge	Beginners; testing; strict budgets	Most people who want balance

Note: In real life, bankroll paths jump. Even good systems can see deep dips. A drawdown is normal. Your plan should expect it.

Implementation, Tools, and Discipline

Define your bankroll. Only use money you can afford to lose. Keep it separate from life needs.
Pick a system. If you are new, pick flat or a small percent (like 1%). If you have a tested edge, try 0.25–0.5 Kelly.
Set rules. Write your unit or percent. Set a max stake per bet. Do not break your rules mid-run.
Track everything. Keep a simple log: date, market, odds, stake, close price, result, and notes. A basic spreadsheet works well. See concepts like EV and standard deviation to understand swings.
Review often. Every 50–100 bets, check if your edge looks real. If not, size down.
Mindset. No chasing losses. No “double to get even” (martingale is risky). Avoid the gambler’s fallacy.
Pick trusted books. Odds quality, limits, and payout speed matter. Compare fair options before you stake. Independent reviews at https://norskecasinoguiden.com/ can help you see which sites have good service, clear terms, and fast withdrawals. Do your own checks too.

For legal info in the UK, see the Gambling Commission. For general math help, see Khan Academy: Statistics and Probability. For a simple Kelly refresher, see the Kelly overview.

Common Pitfalls and Myths

Overestimating your edge: This is the #1 error. Be strict. Small sample wins do not prove skill. Use large samples.
Chasing losses: Doubling down feels good but often ends bad. Stick to your plan size.
Correlated bets: If your picks move together, your risk is higher than you think. Size down for linked bets.
Using full Kelly with weak data: If your p is shaky, use fractional Kelly or a percent plan.
Ignoring fees and limits: Some books cap stakes or move lines fast. Plan for this.
Thinking sizing can fix no edge: No system can save a negative EV strategy.

How to Choose a Bet Sizing System

If you are new or not sure you have an edge: Use flat betting or 0.5%–1% percentage. Focus on learning, tracking, and self-control.
If you have some edge but it is not stable yet: Try 1%–2% percentage or 0.25–0.5 Kelly. Use caps. Review every 100 bets.
If you have strong, tested edge and a long horizon: Consider fractional Kelly (like 0.5). Only move up after real, long-term proof.

Make a simple flow rule: “If variance is high or edge is unclear, size down. If variance is low and edge is strong, size up a bit.” Always err on the safe side.

Responsible Gambling and Legal Notes

Follow the legal age rules (often 18+ or 21+). Check local law before you bet.
Set a budget and time limit. Never bet with money for rent, food, or bills.
If you feel stress, stop and take a break. If you cannot stop, seek help: BeGambleAware (UK) GamCare (UK) National Council on Problem Gambling (US)
BeGambleAware (UK)
GamCare (UK)
National Council on Problem Gambling (US)
This article is educational. It is not betting or financial advice.

BeGambleAware (UK)
GamCare (UK)
National Council on Problem Gambling (US)

FAQs

Is the Kelly Criterion good for sports betting?

Kelly is good if you can estimate true win chance well. If your estimate is off, full Kelly can overbet. Many people use half-Kelly or less. Learn more at Wikipedia and in the MacLean–Thorp–Ziemba review.

How many units or what percent should I bet per game?

There is no one right number. If you are new, 0.5%–1% per bet keeps swings small. If your edge looks real, 1%–2% can work. Kelly gives a math answer, but use fractional Kelly if you are not sure about your estimates.

Is flat betting profitable?

Flat betting does not make a losing strategy win. You still need an edge. Flat betting is good for discipline and learning. It keeps risk simple while you test your picks.

What is fractional Kelly and why use it?

Fractional Kelly means you bet a fraction of the Kelly stake (like 1/2 or 1/3). It cuts drawdowns and pain if your edge guess is wrong. It is a common, safer tweak.

Does bet sizing matter if I don’t have an edge?

It matters for risk control, but it cannot beat the house long-term without edge. Focus on fair odds, good info, and strict rules. Do not expect profit from sizing alone.

How do I estimate my edge in a simple way?

Track your bets over a big sample. Compare your picks to the closing line and market price. Study basics of EV on Khan Academy. Be strict and reduce bias.

Can I mix systems?

Yes. Many people use percent staking day to day and switch to fractional Kelly only when a model shows a strong edge. Keep rules clear to avoid confusion.

Two More Worked Examples

Flat vs Percentage on a Short Run

Setup: Bankroll $1,000. You place 10 even-odds bets. You win 6, lose 4 (this is a good mini-run).

Flat $10 per bet: Profit = (6 × $10) − (4 × $10) = $20. End roll = $1,020.
2% percentage: Stakes change a bit each time (start $20, then adjust). End roll will be a bit higher than flat because stake rose after wins. But if the run was bad, percentage would have cut loss by shrinking stake.

Both are simple. Percentage reacts to results; flat does not.

Kelly with Edge Error

Setup: Odds 2.50 (b = 1.50). You think p = 0.45. Kelly says f = (1.50 × 0.45 − 0.55) / 1.50 = (0.675 − 0.55)/1.50 = 0.125/1.50 ≈ 0.0833 → 8.3% stake.

If the true p is only 0.42, the real Kelly stake should be:

f = (1.50 × 0.42 − 0.58) / 1.50 = (0.63 − 0.58)/1.50 = 0.05/1.50 ≈ 3.3%.

Big difference. This is why many people use half-Kelly or less unless they are very sure.

Practical Tips to Boost E-E-A-T (Trust and Quality)

Be transparent: State your method. Show a few worked bets with math in your log.
Use sources: Read and cite primary or expert works, like Kelly (1956), Thorp, and academic reviews.
Keep content fresh: Review your staking rules every few months.
Protect readers: Use clear warnings against risky behavior. Link to help groups.

References and Further Reading

Kelly, J. L. Jr. (1956). A New Interpretation of Information Rate (original paper).
Thorp, E. O. Official site and writings.
MacLean, Thorp, Ziemba. The Kelly Capital Growth Investment Criterion: Theory and Practice.
Wikipedia: Kelly Criterion (overview and math).
Wikipedia: Expected Value (EV basics).
Wikipedia: Gambler’s Ruin (risk of ruin idea).
Khan Academy: Statistics and Probability (free lessons).
UK Gambling Commission (regulation and player info).
BeGambleAware and GamCare (help and tools).

About the Author and Editorial Note

Author: A bankroll and risk student who tests staking plans with real data and simple math. This piece was reviewed for clarity and accuracy (terms, formulas, and examples checked against the sources above). Last updated: [Insert date].

Editorial standards: We aim for clear, safe, and useful guides. We avoid hype and “sure things.” We add sources and simple examples so you can check the logic yourself.