Bankroll Management for Sports Betting: Kelly Stake Sizing
Size each sports bet from edge and odds with the Kelly criterion, and see why Edward Thorp kept estimates conservative in his $50,000 field test.

Photo: Jakub Żerdzicki on Unsplash
Jump to a section
- 01What the Kelly criterion says about stake size
- 02How Edward Thorp sized real sports bets with Kelly
- 03Why fractional Kelly beats full Kelly when your edge is uncertain
- 04Putting Kelly stake sizing into practice
- 05Guide questions
- How do you manage a gambling bankroll on sports bets?
- Is it possible to be a profitable sports bettor?
- Why bet less than the full Kelly stake?
- What happens if you overestimate your edge?
- 06Sources and further reading
Bankroll management for sports betting means sizing every stake from two inputs, your edge and the odds, instead of from habit or mood. The Kelly criterion turns those inputs into the stake that grows a bankroll fastest over the long run, and fractional Kelly (betting a fixed share of that stake) is the safer choice whenever your edge is an estimate, which in sports it always is. No staking plan makes a losing bet win: it only controls how much damage you take while you find out whether your edge is real.
What the Kelly criterion says about stake size
The Kelly stake grows with your advantage and shrinks with the payout: the bigger your expected profit per unit staked, the larger the share of bankroll you put down, and with no edge the Kelly stake is zero.
For a simple win-or-lose bet the rule fits in one line. Kelly fraction = edge ÷ net odds, where edge is your expected profit per unit staked and net odds are the profit a winning bet returns per unit staked. Two consequences follow, and both matter more than the formula itself:
- Stakes are a share of bankroll, not a fixed sum. After a win the next stake rises with the bankroll; after a loss it falls, which slows the bleeding in a bad run.
- The stake is only as good as the edge you feed in. The odds are known exactly. Your edge is a guess, and Kelly treats that guess as if it were true.
That second point is where most betting bankroll management goes wrong, and it is why the rest of this guide is about scaling the Kelly stake down.
How Edward Thorp sized real sports bets with Kelly
Edward Thorp, the mathematician best known for card counting, ran a sports betting field test in 1993 and used Kelly to size the bets, starting from a $50,000 test bankroll. It is one of the few published records of Kelly staking applied to actual sports wagers.
| Detail of the 1993 test | What Thorp reported |
|---|---|
| Starting bankroll | $50,000, set aside for the test |
| Staking method | Kelly system used to size each bet |
| Typical expectation per bet | About 6% of the amount wagered |
| Deepest dip into the bankroll | About $10,000, roughly a fifth of the test bankroll |
| How edges were estimated | Deliberately conservative |
Thorp also explains why he kept his estimates low on purpose: in his words, overbetting is much more severely penalized than betting too small. That asymmetry is the whole case for fractional Kelly.
Why fractional Kelly beats full Kelly when your edge is uncertain
Fractional Kelly stakes a set share of the full Kelly amount, giving up some growth in exchange for much smaller swings, and it is the standard fix when you cannot be sure of your edge.
The reasoning is laid out plainly in a recent arXiv paper: in real gambling you may have overestimated the edge, so it is safer to bet some fraction of the Kelly stake, at the cost of a slower growth rate. A paper by Hsieh, Barmish and Gubner sums up the wider literature the same way, noting that Kelly bets are widely judged too aggressive on risk measures and that scaling down the bet size is the usual remedy, known as Fractional Kelly Betting.
Here is how the choices compare when the edge you plug in might be wrong.
| Staking approach | How the stake is set | Long-run growth | Swings and drawdowns | If your edge is overestimated |
|---|---|---|---|---|
| Full Kelly | Edge ÷ net odds, as a share of bankroll | Highest, if the edge is exactly right | Large | You are overbetting, the expensive side of the error |
| Fractional Kelly (half, quarter) | A fixed share of the full Kelly stake | Somewhat lower | Much smaller | A cushion absorbs much of the mistake |
| Flat stake | The same sum every bet, ignoring edge | Not tied to your edge | Depends on the sum chosen | No link to edge, so no protection either way |
A fraction of Kelly turns a confident but uncertain estimate into a stake you can survive being wrong about.
Putting Kelly stake sizing into practice
A workable routine has four parts, and none of them requires a betting system that claims to win.
- Estimate your edge, then trim it. Use the most cautious figure you can defend, as Thorp did, since the mistake you can least afford is overstating it.
- Convert it to a full Kelly stake. Divide the edge by the net odds and apply that share to your current bankroll. No edge means no bet.
- Scale it down. Bet a fixed fraction of the Kelly stake, such as half or a quarter, and keep that fraction consistent.
- Resize as the bankroll moves. Recalculate the stake from the current balance, not the starting one, so losing runs automatically shrink your exposure.
To run the numbers for flat, fractional and Kelly staking side by side, use our bankroll management calculator.
Guide questions
How do you manage a gambling bankroll on sports bets?
Size each stake from your estimated edge and the odds rather than by feel, estimate that edge conservatively, and bet a fraction of the full Kelly stake. Recalculate the stake as the bankroll grows or shrinks, and pass on bets where you have no edge at all.
Is it possible to be a profitable sports bettor?
Only with a genuine edge, and even then nothing is guaranteed. Edward Thorp's 1993 test worked with a typical expectation of about 6% of the money wagered, and he still set his estimates low because overbetting is punished far harder than underbetting.
Why bet less than the full Kelly stake?
Because the Kelly stake is only as good as your edge estimate. Researchers note that an edge can easily be overestimated, so a fraction of Kelly is safer, and the price is a slower rate of bankroll growth.
What happens if you overestimate your edge?
You end up staking more than your real advantage justifies, which is the costly side of the mistake. Thorp's point is that too large a bet is penalised much more severely than too small a bet, so a margin of caution in the estimate pays for itself.
Sources and further reading
- Edward O. Thorp (Handbook of Asset and Liability Management, ch. 9), The Kelly Criterion in Blackjack, Sports Betting, and the Stock Market, accessed October 2026
- arXiv, The Kelly Criterion And Utility Function Optimisation For Stochastic Binary Games: Submartingale And Supermartingale Regimes, February 24, 2025
- arXiv (Hsieh, Barmish, Gubner), Kelly Betting Can Be Too Conservative, October 4, 2017
Topic of this guide
Pass it on
Choosing where to play? See the casinos we cover and our take on each.
Where to play



