Bankroll Management for Sports Betting — Kelly, Flat & Percentage Methods

Bankroll management is the part of sports betting that nobody finds exciting, and the part that determines whether anyone ever survives long enough to find out whether they have a real edge. The mathematical truth is direct: even a bettor with a real long-term edge will lose all their money over time if they stake too aggressively. And a bettor with no edge will lose their money regardless of staking — but they will lose it faster with aggressive staking.

This guide covers the three standard staking methods used across professional and informed recreational betting: flat staking, percentage of bankroll, and the Kelly criterion. We explain why Martingale and progressive doubling-up systems are mathematically guaranteed to bankrupt you eventually, and we explain how to track ROI honestly rather than fooling yourself with wins-and-losses summaries. We do not select bets and we do not predict outcomes.

Step zero — define your bankroll honestly

Your bankroll is the amount of money you have mentally written off as entertainment spend. It is not your savings. It is not money you might want for rent next month. It is not borrowed money. It is not money you can’t tell your partner about. It is a defined sum that, if you lost it entirely tomorrow, would not change your life.

For most recreational bettors, this should be a few percent of monthly disposable income at most. The UK Gambling Commission’s Responsible Gambling Code and similar codes elsewhere consistently frame gambling as an entertainment cost, not as financial investment.

If you cannot honestly state your bankroll number out loud right now, you do not have a bankroll. You have a leaking account. Stop reading and define the number first.

Method 1 — Flat staking

Flat staking is the simplest and most honest method for recreational bettors. Every bet is the same size, expressed as a percentage of bankroll.

The standard recommendation: 1% of bankroll per bet. Aggressive: 2%. Conservative: 0.5%.

Worked example

Bankroll £1,000. Flat stake at 1%.

Every bet: £10
After 50 bets, win 25 / loss 25 at average decimal 2.00: you are flat (50 × £10 = £500 staked, 25 winners × £20 = £500 returned).
After 100 bets, win 48 / loss 52 at average 2.00: you are down £40 (lost 52 × £10 = £520, won 48 × £10 = £480 net). Standard variance.
After 500 bets, even with a 1% structural disadvantage from bookmaker margin, your worst-case drawdown stays manageable.

The single biggest advantage of flat staking is psychological: you do not have to make a decision about stake size for each bet. The stake is fixed. The only decision is whether to bet at all.

Adjusting flat stake over time

If your bankroll grows from £1,000 to £1,300, you can recalculate your 1% to £13 per bet. If it shrinks to £700, your 1% is £7. This is technically the simplest “percentage of bankroll” variant, but most flat-staking bettors recalculate weekly or monthly rather than after every bet.

Method 2 — Percentage of bankroll

Percentage staking is mechanically the same as flat staking adjusted after every bet rather than on a calendar cycle. It is mathematically equivalent to compounding — you bet more after wins and less after losses, automatically.

Worked example

Bankroll £1,000. Stake 2% of current bankroll every bet.

Bet 1: stake £20.00 at 2.00. Win → bankroll £1,020.
Bet 2: stake £20.40 (2% of £1,020). Lose → bankroll £999.60.
Bet 3: stake £19.99 (2% of £999.60). Lose → bankroll £979.61.
Bet 4: stake £19.59. Win → bankroll £999.20.

The advantage: automatic capital protection during losing streaks. The disadvantage: slower growth during winning streaks compared to fixed stake.

This is the default sane choice for any informed recreational bettor. The 1–2% range is robust across most realistic edge sizes and variance profiles.

Method 3 — Kelly criterion

The Kelly criterion is the mathematically optimal staking formula for long-term geometric growth of bankroll, assuming you know your true edge. It was originally derived in 1956 for information theory and applied to gambling by Ed Thorp.

The formula

f* = (b × p − q) ÷ b

Where:

f* = fraction of bankroll to stake
b = decimal odds − 1 (the profit multiplier, e.g. 2.50 → 1.50)
p = your estimated probability of winning
q = 1 − p (your estimated probability of losing)

Worked example

You estimate Liverpool has a 60% probability of beating Brighton. The decimal odds are 2.00.

b = 2.00 − 1 = 1.00
p = 0.60
q = 0.40

f* = (1.00 × 0.60 − 0.40) ÷ 1.00 = 0.20

Full Kelly says: stake 20% of your bankroll on this bet.

That is a lot. If your bankroll is £1,000, you would stake £200 on one football match. Most bettors find this terrifying, and they are right to.

Why full Kelly is too aggressive for almost everyone

Full Kelly is mathematically optimal only if your probability estimate is exactly correct. In real betting, your estimate is almost certainly slightly wrong. If you think Liverpool’s probability is 60% and the true probability is 55%, full Kelly bets are not just suboptimal — they are aggressively negative-EV and will bankrupt you fast.

The fix is fractional Kelly:

Half Kelly: stake 50% of what the full Kelly formula recommends. Reduces growth rate but dramatically reduces variance and ruin risk.
Quarter Kelly: stake 25% of full Kelly. Conservative.
One-eighth Kelly: stake 12.5% of full Kelly. Used by very disciplined professional bettors.

For the example above, quarter Kelly would be 5% of bankroll instead of 20%. That is still aggressive by recreational standards but at least defensible.

When NOT to use Kelly

Kelly assumes you can quantify your true edge. If you cannot honestly quantify your edge — and most recreational bettors cannot — Kelly outputs garbage. The formula will tell you to stake huge amounts on bets where you think you have a big edge but actually do not.

Practical rule: use Kelly only if you also track closing line value (CLV) and have demonstrated positive CLV over hundreds of bets. Without that demonstrated track record, Kelly is a way to lose money faster than any other system. See our value betting guide for more on CLV.

Martingale — the system that always bankrupts

Martingale is the staking system that says: after every loss, double your stake; after every win, return to base stake.

The theory: eventually you will win, and the win pays back all previous losses plus one unit of profit.

The reality: every staking system that ignores bankroll constraints fails when the bankroll runs out, and Martingale runs out very fast.

Worked example — bankruptcy in 10 spins

Starting stake £10 on a 50/50 bet at decimal 2.00. You lose 10 in a row.

Bet #	Stake	Cumulative loss
1	£10	£10
2	£20	£30
3	£40	£70
4	£80	£150
5	£160	£310
6	£320	£630
7	£640	£1,270
8	£1,280	£2,550
9	£2,560	£5,110
10	£5,120	£10,230

After ten losses in a row — which happens once every 1,024 sequences on a true 50/50 — you have either bankrupted yourself or hit the sportsbook’s maximum stake cap (most sportsbooks cap individual stakes at £5,000–£10,000 on most markets).

The probability of ten consecutive losses on a coin flip is 1 in 1,024, which sounds rare. But Martingale assumes you can survive any losing streak, no matter how long. Over thousands of betting sequences, ten-loss streaks happen. The first time one happens on the same evening you raised your base stake by 50%, you’re done.

Every staking system that scales up after losses has the same flaw. Fibonacci, D’Alembert, Labouchère — all variants of “increase after a loss”. All carry exactly the same mathematical guarantee: ruin in finite expected time, given any non-zero margin and any bankroll constraint.

The mathematical proof of this is straightforward and has been covered in betting and gambling literature for decades. We do not recommend any progressive doubling system.

Track ROI, not wins

The most common self-deception in recreational betting is tracking wins and losses instead of ROI.

A 60% win rate sounds great until you realise:

60% win rate at average decimal odds 1.40 → ROI of (0.60 × 0.40) − (0.40 × 1.00) = −16%. Losing money.
60% win rate at average decimal odds 1.80 → ROI of (0.60 × 0.80) − (0.40 × 1.00) = +8%. Winning.
40% win rate at average decimal odds 3.00 → ROI of (0.40 × 2.00) − (0.60 × 1.00) = +20%. Strongly winning.
90% win rate at average decimal odds 1.10 → ROI of (0.90 × 0.10) − (0.10 × 1.00) = −1%. Losing money despite “winning almost every bet”.

Win rate alone is meaningless. The ROI formula is:

ROI = Total profit ÷ Total stake

Track every bet in a spreadsheet with stake, odds, outcome, profit, and a notes column. Calculate ROI monthly. Numbers under +2% over 500+ bets are essentially break-even given variance. Numbers above +5% across 1,000+ bets at sensible margins suggest real edge — though most “real edge” bettors get account-limited by sportsbooks before they can scale this up.

Stop-loss and stop-win — soft tools that matter

Beyond the mathematical staking method, two soft tools help most recreational bettors:

Stop-loss: set a daily or weekly loss limit. When hit, stop. Do not place “one more bet to win it back”. Loss chasing is the most documented behaviour in problem gambling research — see GamCare’s guidance.

Stop-win: set a level at which you walk away from a winning session. Sounds counterintuitive — “but I’m running hot, why stop?” — but human psychology in losing-back hot streaks is statistically grim. Bettors on winning streaks tend to push stakes up, run into variance, and finish below where they would have if they had stopped at the soft limit.

Both stop levels work best when set in advance and not adjusted in the moment.

Sportsbook tools to enforce discipline

Every UKGC-licensed sportsbook is required to offer:

Deposit limits (daily / weekly / monthly)
Time-out periods (24-hour to 6-week voluntary cool-off)
Self-exclusion (6-month to 5-year permanent)

Most other regulated markets — Ontario AGCO, NZ TAB, IE Revenue (GRAI), AU NCPF — require similar tools.

These tools work better when set at the start of betting, before any urgency feeling. A £200/week deposit limit set during a calm Sunday afternoon is much more effective than one set after losing a Premier League match on Saturday night.

The UK’s GAMSTOP self-exclusion register covers all UKGC-licensed operators in one place. The equivalent in Australia is BetStop (national since 2023). These exist for a reason. Use them before you need them.

A worked example of a full bankroll plan

Here is what a realistic recreational bankroll plan looks like for someone with £1,000 in entertainment-spend bankroll:

Element	Decision
Bankroll	£1,000 — defined and ring-fenced from savings
Staking method	Flat percentage, 1% of current bankroll, recalculated weekly
Base stake	£10 per bet
Bet types	Singles only on main markets (lowest margin), no accumulators, no specials
Sportsbook count	2 books with deposit limits set at £50/week each
Weekly stop-loss	£100 total across both books
Time-out	Auto 24-hour cool-off after any single day with £50+ loss
Tracking	Spreadsheet with every bet, ROI calculated monthly
Bankroll review	Monthly: re-set 1% based on current balance, withdraw any growth above £1,200

This is not exciting. It is also not where bankrolls die. The recreational bettors who survive long enough to ask whether they have an edge tend to look like this. The ones who burn out tend to skip every element on this list.

Frequently asked questions

What is the safest staking method for beginners?

Flat percentage staking at 1% of bankroll per bet. Mechanically simple, naturally limits damage during losing streaks, automatically compounds during winning streaks.

Does Kelly work in practice?

For professional bettors who can accurately quantify edge, fractional Kelly (quarter or half) is a standard tool. For recreational bettors who cannot quantify edge, Kelly produces aggressive stake sizes that compound losses fast. Most bettors should not use Kelly.

Why is Martingale always wrong?

Because it ignores bankroll constraints and sportsbook stake caps. Any sufficiently long losing streak — and they happen — exhausts the bankroll or hits the cap. The “guarantee” of eventual win only holds with infinite bankroll and no stake limits, neither of which exists in reality.

How big should my bankroll be?

As big as you can afford to lose entirely without changing your life. For most recreational bettors, that is a few hundred to a few thousand pounds, replenished only from genuine entertainment-budget surplus.

Should I track every bet?

Yes, if you want to know whether you have an edge. Every winning bettor tracks ROI per market segment. Untracked betting cannot be analysed.

Can I use the same bankroll across multiple sportsbooks?

Mentally yes, accounting-wise yes — just track the total across all books and apply your stake-sizing rule to the total. Most professional bettors hold accounts at 5–15 books to access price-shopping and to spread account-limit risk; recreational bettors typically use 2–3.

Responsible gambling — the most important section

Bankroll management is a survival tool, not a profit tool. It buys you time, not edge. It limits damage during normal variance, but it does not change the structural negative expectation built into bookmaker margins.

The honest hierarchy:

The math says most bettors lose money long-term because bookmaker margins compound across thousands of placements.
Bankroll management determines how long that loss takes, not whether it happens.
Real edge is rare, hard to verify in fewer than thousands of bets, and difficult to scale because sportsbooks limit winners.
Most “winning strategies” sold online are statistical illusions, survivorship bias, or outright fraud — UK ASA has ruled against many of them.

If your betting is consuming more attention than entertainment justifies, if you are betting more than your defined bankroll allows, or if you are chasing losses, free confidential support is available:

UK: BeGambleAware.org · GamCare 0808 8020 133 · GAMSTOP self-exclusion
Ireland: problemgambling.ie
Canada (Ontario): ConnexOntario 1-866-531-2600
New Zealand: Gambling Helpline 0800 654 655 · BetStop register
Australia (when applicable): Gambling Help Online 1800 858 858 · BetStop national self-exclusion

Sports betting outcomes are statistically uncertain. Bankroll management does not change that. It only changes how quickly losses accumulate when variance turns against you. We do not offer tips, picks, or selections. We explain math.