Kelly Criterion Formula Explained: Inputs, Edge, and Fractional Kelly

The Kelly Criterion formula is a simple rule that tells you how much to bet when you have an edge. It is not a magic trick. It is a sizing tool that tries to grow your money over time while avoiding ruin. Many people in betting and trading use it to guide position size.

If you bet too much, you can lose fast even if the odds are good. If you bet too little, you leave growth on the table. The Kelly Criterion gives a clear number between 0% and 100% of your bankroll for each bet. This number depends on your estimate of the win rate and the odds.

In this guide, you will learn what the inputs mean, how to read the formula, how to estimate your edge, and how to use fractional Kelly to control risk. We will use short sentences, simple words, and no jargon. We also include two tables for quick review. By the end, you will know how to apply the Kelly Criterion formula in a careful and human way.

The Kelly Criterion Formula, Explained Simply

The Kelly Criterion formula uses three core inputs: your chance to win (p), your chance to lose (q), and the net odds (b). It outputs a stake f* that is a fraction of your bankroll. The formula is:

• f* = (b*p − q) / b

The output reacts in intuitive ways:

• If your edge goes up, f* goes up.

• If odds pay more (higher b) for the same p, f* goes up, but not linearly.

• If your p drops, f* drops and can become zero or negative.

Keep in mind that b must reflect the true net odds. With decimal odds O, b = O − 1. If you use American odds, convert to decimal first, then subtract 1.

What Is the Kelly Criterion?

The Kelly Criterion is a rule for optimal bet sizing. It comes from a math idea called logarithmic growth. The goal is not to win every bet. The goal is to grow your bankroll at the highest long-term rate while keeping a nonzero chance of survival. The Kelly Criterion does this by telling you the fraction of your bankroll to stake when you have a positive edge.

You only use Kelly when you believe you have positive expected value (also called “positive EV”). If your edge is zero or negative, the Kelly stake is zero. This is key. Kelly is not a system that turns a losing bet into a winning one. It is a sizing rule for good bets.

Kelly works in many fields. Sports bettors use it when the odds on the board differ from their own fair odds. Poker players use it when they know their chance to win a hand against a range. Some traders use a version of Kelly to set leverage or position size when they estimate return and risk. The core idea is the same: stake in proportion to your edge, not your mood.

Inputs and Edge: How to Get p Right

Getting p right is the core job. Use data, not hope. Clean your data, include costs, and be conservative. Test your method out of sample. If your skill is early and uncertain, use Fractional Kelly.

You can also widen your view of “edge.” Sometimes your raw p is only one part. Maybe you can get better lines than the market, or you can time your entry in trading to reduce costs. These also improve EV and thus f*.

When your estimates vary over time, do not swing sizes wildly. For example, if p moves from 0.55 to 0.57 with little new data, hold your size steady or adjust only slightly. Stability helps you stick to the plan.

The Kelly Criterion Formula and Its Inputs

The most common form of the Kelly Criterion formula for a single bet is:

• f* = (b * p − q) / b

where:

• f* is the fraction of your current bankroll to bet (between 0 and 1; think 0% to 100%).

• p is the probability you win the bet (your estimate).

• q is the probability you lose the bet, so q = 1 − p.

• b is the net odds you win on a success.

What is “net odds” b?

If you use decimal odds O (like 2.50), then b = O − 1 (so 2.50 odds means b = 1.50). If you use American odds or fractional odds, convert them to net odds first. The “net odds” means how much you win net on a 1-unit stake, not counting your own stake.

Special case: even-money bets

If b = 1 (even money), the formula becomes:

• f* = p − q = 2p − 1

So if you think you win 55% of even-money bets (p = 0.55), then f* = 0.10, or 10% of bankroll.

Interpretation

• If f* is positive, you have edge.

• If f* is zero or negative, your edge is zero or negative, so you should not bet.

Kelly Criterion Inputs and Meanings

Item	Symbol	What it means	Typical range	How to estimate (simple)
Win probability	p	Chance the bet wins	0 to 1	Model, data, or expert judgment
Lose probability	q	Chance the bet loses	0 to 1	1 − p
Net odds	b	Net profit per 1 unit if win	> 0	From market odds; for decimal odds O, b = O − 1
Kelly fraction	f*	% of bankroll to stake	0 to 1 (often small)	Compute (b*p − q) / b
Edge (EV per unit)	EV	Expected value of 1-unit bet	Any real number	EV = b*p − q
Bankroll	B	Money set aside for this game or strategy	> 0	Your choice;keep it separate fromlife funds

A simple check: if EV = b*p − q ≤ 0, then f* ≤ 0 and you skip the bet. If EV > 0, the sign is green, but size still matters. Kelly helps you scale the bet by both the size of your edge and the payoff odds

Also Read: What is the Black–Scholes Formula? How It Helps Value Options

Estimating Edge for the Kelly Criterion Formula

The Kelly Criterion formula depends on p (and thus q) and b. The odds b are known from the market. The hard part is p. If you overestimate p, you will bet too much. So you must estimate p with care.

Here are simple ways to estimate p:

1. Historical Frequency

• Look at a large set of past, similar events.
• Use a clean sample (same league, same type of bet, similar conditions).
• Compute win rate.
• Adjust for changes that matter (injuries, rules, team form).
• The more data you have, the more stable the estimate.

2. Model-Based Estimate

• Build or use a model that predicts outcomes and probabilities.
• For sports: use team strength, player ratings, pace, luck factors.
• For trading: use signals that predict returns; estimate probability a trade is profitable after costs.
• Validate on past data not used to train the model.

3. Market-Implied Probability

• Convert the offered odds to implied probability (after removing the bookmaker margin).
• Compare with your model probability.
• If your p is higher than market-implied p, and you trust it, you may have positive edge.

4. Bayesian Update

• Start with a prior belief (e.g., “I think this team wins ~40%”).
• Add new evidence (injury news, matchups).
• Move your estimate but do not overreact.
• This helps avoid wild swings from small news.

5. Cost and Slippage

• In trading, include fees, spreads, and slippage.
• In betting, include the bookmaker margin.
• Edge after costs is what matters.

Rule for caution: When in doubt, reduce p a bit or use fractional Kelly (explained below). Estimation error is the main risk in Kelly use.

Fractional Kelly: Why and How

Full Kelly uses the raw formula. It maximizes long-run growth if your p and b are correct. But in real life, your p is noisy. Full Kelly can create large swings and deep drawdowns. Many users prefer Fractional Kelly to reduce risk.

Fractional Kelly rule:

• f(λ) = λ × f*, where 0 < λ ≤ 1.

Common choices are λ = 0.5 (Half Kelly) or λ = 0.25 (Quarter Kelly). Fractional Kelly reduces both expected growth and volatility. The benefit is smoother equity and less damage from p estimation errors.

Why fractional Kelly helps:

• Estimation error: If you overstate p, full Kelly overbets. A fraction cuts the mistake.
• Fat tails: Real returns can be lumpy. A smaller f means smaller hits in bad runs.
• Human comfort: Many people quit after a big drawdown. Fractional Kelly helps you stick to the plan.

How to pick λ:

• Start with Half Kelly if you are new.
• If your model is young or noisy, use Quarter Kelly or less.
• If your model is proven and stable, you may move toward Full Kelly, but do so slowly and with logs of your results.

Fractional Kelly Comparison (Qualitative)

Kelly level	Stake (vs. Full)	Expected growth	Volatility of bankroll	Typical use case
Full Kelly (λ = 1.0)	100%	Highest if p is correct	Highest	Proven model, strong edge, high tolerance
Half Kelly (λ = 0.5)	50%	~75%–80% of Full (often)	Much lower	Balanced growth and risk
Quarter Kelly (λ = 0.25)	25%	~50%–60% of Full (often)	Lower still	New model, high uncertainty
Eighth Kelly (λ = 0.125)	12.5%	Lower	Very low	Very high uncertainty or long learning phase

These growth percentages are rough, not exact. The point is simple: you give up some growth to gain stability.

Practical Rules for Using the Kelly Criterion Formula

• Only bet when EV is positive: Compute EV per unit: EV = b*p − q. If EV ≤ 0, do not bet. This avoids “action for action’s sake.”

• Keep a clean bankroll: Your bankroll should be money set aside for this method only. Do not mix with rent money. Adjust the bankroll level only at set times (e.g., weekly) to avoid overreacting to swings.

• Use Fractional Kelly by default: Unless your model is very strong, start at Half Kelly or less. You can increase later with evidence.

• Respect correlation: If two bets are linked (same team, same event, same market), do not treat them as independent. Either treat them as one combined bet or reduce size more.

• Cap the stake: Even if Kelly suggests a high fraction, set a hard max (e.g., 5% of bankroll). This protects against model error and black swan events.

• Re-estimate p often, but not too often: Update your model with new data on a schedule. Avoid changing p wildly from short streaks.

• Track results: Keep a log: date, bet, p, odds, stake, result, and bankroll. Over time, compare realized results vs. your p. Calibrate. This improves your p and your use of Kelly.

Also Read: What Is Risk-Neutral Probability? Theory, Models, and Applications

Common Mistakes and How to Avoid Them

Mistake 1: Overconfidence in p.

People tend to be too sure. If you inflate p, Kelly overbets. Fix: use a lower p or Fractional Kelly. Add a margin of safety.

Mistake 2: Ignoring costs.

Fees, spreads, and book margins reduce EV. Fix: include all costs in your EV and in your p estimate.

Mistake 3: Chasing losses.

Kelly never says “double to catch up.” It says “size by edge and bankroll.” Fix: follow the formula, not emotions.

Mistake 4: Overlapping bets.

Many bets on the same game act like one big bet. Fix: treat them as one and size once. Or reduce each stake.

Mistake 5: Overbetting long shots.

High odds can tempt you. But if p is small and uncertain, Kelly can still be small. Fix: be strict with p; if unsure, size down.

Mistake 6: Using Kelly with zero edge.

No sizing rule can save a losing game. Fix: first, find an edge; second, size with Kelly.

Risk, Drawdowns, and Time

Kelly aims for long-run growth, not short-term comfort. Drawdowns still happen. Even with positive EV, you will face losses. A key lesson: with Full Kelly, drawdowns can be large. Fractional Kelly reduces this pain.

A useful mental model:

If you bet Full Kelly on even-money with p = 0.55 (f* = 10%), a 5-loss streak cuts your bankroll by about:

• After 1 loss: 0.90

• After 2 losses: 0.90² = 0.81

• After 5 losses: 0.90⁵ ≈ 0.5905 → ~41% drawdown

That is heavy. With Half Kelly (5%), the same 5-loss streak is 0.95⁵ ≈ 0.7738 → ~22.6% drawdown.

These are simple, not perfect, but they show the trade-off. Smaller f means smaller hits and easier recovery.

Time to recover: Recovery is nonlinear. A 50% drop needs a 100% gain to get back. By keeping f moderate, you shorten the path back after a bad run. This is one more reason many users favor Half or Quarter Kelly.

Conclusion

The Kelly Criterion formula is a clear way to size bets when you have an edge. It tells you what fraction of your bankroll to stake based on your win chance and the odds. If EV is not positive, you do not bet. If EV is positive, you size in proportion to that edge.

In real life, your p is noisy. Full Kelly can be rough. Fractional Kelly—like Half or Quarter—gives smoother growth and helps you stick with the plan. Add guardrails, track your results, and keep your bankroll separate from life money. Over time, improve your estimates and your process.

Use Kelly with care and respect. It will not fix a losing game, but it can help a winning one grow in a steady way. Keep things simple, measure your edge, and size with discipline. That is how the Kelly Criterion formula serves you best.