Kelly Criterion vs. Flat Betting: Bankroll Strategies

The Kelly criterion sizes each bet by your edge, betting more when the odds undervalue your pick and nothing when they don't, while flat betting stakes the same amount on every bet. Kelly grows a bankroll faster if your win probabilities are accurate, but it swings hard and punishes overconfidence. Flat betting grows more slowly and forgives mistakes. Most bettors are better off flat betting 1% to 2% of their bankroll, and bettors with a tested model usually use a fraction of Kelly rather than the full amount.

Kelly criterion vs flat betting at a glance

What is the Kelly criterion?

The Kelly criterion is a formula for the share of your bankroll to bet so that it grows as fast as possible over the long run. John L. Kelly Jr., a scientist at Bell Labs, published it in 1956, and the mathematician Edward Thorp later applied it to blackjack, sports betting and investing. Its logic is simple: bet in proportion to your edge. No edge means no bet.

The Kelly criterion betting formula

The formula is:

Kelly % = (b × p − q) / b

• b = the net odds, or what you win per $1 staked (decimal odds minus 1). At -110, b is 0.909. At +150, b is 1.5.
• p = your probability of winning.
• q = your probability of losing, which is 1 − p.

An equivalent shortcut is Kelly % = edge ÷ b, where the edge is your expected profit per $1 bet, or p × (b + 1) − 1.

Kelly criterion examples with a $1,000 bankroll

The last row is the important one. At -110 you need to win 52.4% of the time just to break even, so a 52% pick has a negative edge and Kelly says pass. The expected value calculator will show the same thing for any price.

What is flat betting?

Flat betting means staking the same amount on every bet, whatever the odds and however confident you feel. You pick a unit, say $20 from a $1,000 bankroll, and every bet is $20. There are two common versions:

• Flat stake: risk the same amount every time. $20 on a -110 favorite, $20 on a +200 underdog.
• Flat to-win: size each bet to win the same amount. To win $20 you risk $22 at -110 and $10 at +200.

Flat betting's main strength is that it blocks the two habits that drain bankrolls fastest: betting bigger on "sure things" and raising stakes to chase losses. It also gives you a clean record, because every result counts the same. Our guide to unit betting covers how big a unit should be.

How do flat betting and percentage betting compare?

Percentage betting is flat betting that updates. Instead of a fixed $20, you bet a fixed 2% of whatever your bankroll is now. If $1,000 grows to $1,250, the bet rises to $25; if it falls to $800, it drops to $16. That makes losing streaks shrink your risk automatically and winning streaks compound, without needing a probability for each bet. It sits between flat betting and Kelly: same size for every bet, but scaled to the bankroll.

Kelly criterion vs flat betting: what actually happens to your bankroll

With the 55% at -110 example, full Kelly bets $55 on a $1,000 bankroll and a flat bettor using the same starting stake also bets $55. Here is how they differ:

• After five straight losses, full Kelly shrinks each bet as the bankroll falls and ends at $753.63. The flat bettor loses $275 and ends at $725.
• After a hot run, Kelly's bets grow with the bankroll and compound. The flat bettor's profit grows in a straight line.
• When the edge changes, Kelly bets more on a 60% pick than a 55% pick. Flat betting treats them the same.

The catch is the swings. Thorp showed that a bettor using full Kelly has about a 50% chance of seeing the bankroll cut in half at some point, and about a one-in-three chance of halving it before doubling it. Those are the odds for a bettor whose probabilities are exactly right. A bettor who overestimates the edge does worse.

Why full Kelly punishes overconfidence

Kelly trusts your numbers completely. In a well-known experiment by Victor Haghani and Richard Dewey, 61 quantitatively trained players were given $25 to bet on a coin they were told landed heads 60% of the time. Kelly says bet 20% of the bankroll on heads each flip. About 30% of the players went bust anyway, mostly by betting far too much. Sizing, not picking, sank them.

It gets worse when the probability itself is wrong. In the next section, a 3.5-point overestimate turns a 2.4% bet into a 9.7% bet.

Using the Kelly criterion with a football prediction model

Kelly is most useful to bettors who have a football prediction model, because a model produces exactly the input Kelly needs: a win or cover probability for every game. It is also where Kelly does the most damage, because model probabilities are usually more confident than they should be. A sound process looks like this:

1. Get a probability from the model. Say your NFL model gives the Bills a 57% chance to cover at -110.
2. Compare it with the market. Strip the vig out of the market price with a no-vig calculator. A -110/-110 spread implies 50% each side, so your model claims a 7-point edge over the market. Edges that large are rare in NFL sides, which is a warning sign.
3. Shrink toward the market. Markets are hard to beat, so many bettors blend their model with the market price rather than trusting the model alone. A 50/50 blend of 57% and 50% gives 53.5%.
4. Size with fractional Kelly. Full Kelly at 57% says bet 9.7% of the bankroll. At the blended 53.5% it says 2.35%, and quarter Kelly of that is about 0.6%. That spread, from 9.7% down to 0.6%, is why most model bettors cap their stakes.
5. Watch correlated and same-day bets. Kelly assumes bets are independent. A side and an over in the same game, or ten NFL bets all running on one Sunday, are not, so cut each stake when several bets settle at once.
6. Test the model before trusting it. Track whether your 55% picks actually win about 55% of the time, and whether you beat the closing line. If your bets consistently get worse prices than the close, the model's edge is probably not real. The closing line value calculator measures this bet by bet.

For more on building the model itself, see building projection models for betting and how win probability models work in college football.

Fractional Kelly: half Kelly and quarter Kelly

Fractional Kelly means betting a fixed fraction of the full Kelly stake, usually half or a quarter. It is the standard compromise, because the trade between growth and risk is lopsided in your favor:

These figures come from the standard continuous approximation used by Thorp, and they assume your probabilities are correct. Half Kelly keeps three-quarters of the growth for a quarter of the variance. Because real probability estimates contain errors, fractional Kelly also protects you from your own overconfidence: if your true edge is half what you think, half Kelly is in fact full Kelly. For bettors starting out with a model, quarter Kelly is the common choice.

What about the martingale strategy?

The martingale strategy, doubling your stake after every loss so that one win recovers everything, is not a bankroll strategy at all. It is the opposite of Kelly: it raises stakes when the bankroll is shrinking. Starting at $10, seven straight losses cost $1,270, and the eighth bet has to be $1,280 just to win back $10. Losing streaks that long happen to every bettor eventually, and betting limits or an empty bankroll end the system first. At -110 odds, doubling does not even fully recover the losses. Flat, percentage and fractional Kelly staking all avoid that trap by tying bet size to the bankroll, not to the last result.

Which bankroll strategy should you use?

• New or recreational bettor: flat bet 1% to 2% of your bankroll. You do not yet know your edge, and flat staking keeps a clean record while you find out.
• Disciplined bettor without a model: percentage betting, recalculated weekly or monthly, so the stake follows the bankroll.
• Bettor with a tested model or proven closing line value: quarter to half Kelly, with a cap on any single bet and on total same-day exposure.

Whichever you choose, no staking plan turns a losing bettor into a winning one. Staking decides how fast a real edge grows and how much damage a bad run does, so it only helps if the edge exists. A bankroll management tool and an honest look at variance in sports betting help set expectations.

Kelly and flat betting on BettorEdge

Kelly rewards a better price twice: a better number raises your edge, and a bigger edge raises your stake. On BettorEdge, a peer-to-peer marketplace, you bet against other people instead of a house, so there is no vig built into the price, and the same 55% pick at a better number than -110 carries more edge.

The bankroll calculator has a Kelly mode that sizes a bet from the odds and your probability, alongside a simple unit mode for flat bettors. In the app, you can set a default unit size so every ticket's 1x, 2x and 3x buttons fill in the right stake, and every bet is tracked automatically, with ROI by sport and market, so you can check whether your edge is real before you size up. For prediction market traders, see the Kelly criterion for prediction markets.

Kelly criterion vs flat betting FAQ

What is the Kelly criterion in sports betting?

The Kelly criterion in sports betting is a formula that tells you what percentage of your bankroll to stake on a bet, based on your win probability and the odds. The formula is (b × p − q) / b, where b is the net decimal odds, p is your win probability and q is 1 − p. A 55% pick at -110 gives 5.5% of the bankroll, and a negative result means skip the bet.

Is flat betting better than the Kelly criterion?

Flat betting is better than the Kelly criterion for most bettors, because Kelly only works if your win probabilities are accurate and most bettors overestimate their edge. Kelly grows a bankroll faster when the inputs are right, but full Kelly carries about a 50% chance of halving the bankroll at some point. Bettors with a tested model often use half or quarter Kelly instead.

What is fractional Kelly?

Fractional Kelly is staking a fixed fraction of what the full Kelly formula recommends, most often half Kelly or quarter Kelly. Half Kelly keeps about 75% of full Kelly's long-run growth with about a quarter of the variance. It also cushions mistakes, since betting half Kelly on an edge you have overestimated by double is the same as betting full Kelly on your true edge.

How do you use the Kelly criterion with a football prediction model?

To use the Kelly criterion with a football prediction model, take the model's win or cover probability, compare it with the no-vig market price, shrink it toward the market, and then bet a fraction of the resulting Kelly stake. A model that says 57% at -110 implies a 9.7% full Kelly bet, which is far too large to trust. Blending with the market and using quarter Kelly brings it under 1%.

Does the martingale strategy work for sports betting?

No, the martingale strategy does not work for sports betting, because doubling after every loss requires an unlimited bankroll and no betting limits. Starting from $10, seven straight losses cost $1,270 and the next bet must be $1,280 just to win $10. At -110 odds, a single win does not even fully recover the losses.