September 27, 2026·8 min read

Betting bankroll management: unit size and staking

How much to stake per bet: unit size, flat staking, the Kelly criterion done properly and what normal losing runs look like, worked from a real record.

bankrollstakingKelly criterion

Most betting advice spends its time on what to bet. The part that decides whether a real edge survives long enough to pay out is how much you bet — and it is where most bettors quietly go wrong, even with good picks.

This guide covers bankroll management for football betting the way we think about it at Kepler: how big a unit should be, why flat staking beats clever staking for most people, what the Kelly criterion really says, and what a normal losing run looks like when your edge is genuine. The numbers are worked from a record like ours rather than invented.

The short version

• Keep a dedicated bankroll: money set aside for betting that you can afford to lose entirely.
• Stake one unit per pick, with a unit of 1–2% of that bankroll.
• Stake every pick the same — flat staking — and bet each as a single, not an accumulator.
• Expect losing runs of six or more picks. They are normal, not a signal to change anything.
• Judge results over hundreds of bets, and watch closing line value rather than last week's profit.

What a unit is, and how big it should be

A unit is your standard stake, expressed as a share of your bankroll rather than as a fixed amount of money. Everything — profit, drawdowns, ROI — is then measured in units, which is what lets two people with very different bankrolls compare results. Our track record is reported the same way: one unit on every published pick.

You will often see "stake 1–5% per bet". For a realistic edge, the top of that range is far too aggressive. The reason is the size of the losing runs you will meet along the way, which the next two sections put numbers on.

Flat staking: the plan our record uses

Flat staking means betting the same number of units on every pick — one unit, every time — regardless of odds, confidence or how the last few results went.

It sounds unambitious, and that is the point. It is the only plan that is immune to the three habits that sink most bettors: raising stakes to chase losses, raising them after a hot streak, and betting bigger on picks that "feel" stronger. A flat plan makes your results a direct reading of how good the picks are, which is exactly what you want to be able to measure.

Recalculate the size of a unit on a fixed schedule — once a month, say — from the bankroll you actually have. Never mid-way through a losing run.

What losing runs really look like

This is the part most staking guides leave out. Take picks that win about 59% of the time at average odds of 1.80 — close to our published record at the time of writing. That is a genuine edge of about +6% per bet. Here is what flat one-unit staking looks like over 20,000 simulated histories:

Two things follow. First, even a real edge leaves you behind after 100 bets about one time in four. Second, over 500 bets a peak-to-trough drawdown of 14 units is ordinary and 23 units happens one time in ten.

Now convert that into money. With units at 1% of your bankroll, a 23-unit drawdown costs 23% of it — painful, survivable. At 5% it costs well over the whole bankroll. That is why we recommend 1% per unit, 2% at most: not because the edge is small, but because the swings are large.

The Kelly criterion, done properly

The Kelly criterion sizes each bet to maximise long-run bankroll growth, given your edge. With decimal odds o and a win probability p:

Kelly stake = (p × (o − 1) − (1 − p)) / (o − 1)

For a pick with a 60% chance at odds of 2.50, that is (0.60 × 1.5 − 0.40) / 1.5 = 33% of your bankroll. For our kind of pick — 59% at 1.80 — it is 7.75%.

Both numbers are enormous, and there is a good reason not to use them. Kelly assumes you know p exactly. You do not: every probability is an estimate. If the true chance of that 1.80 pick were 55% rather than 59%, the bet would have a slightly negative expected value (−1%) — yet full Kelly would still be staking nearly 8% of your bankroll on it. Overestimating your edge is the normal case, and Kelly punishes it hard.

That is why professionals who use Kelly use a fraction of it. Quarter Kelly on our kind of pick is about 1.9% of bankroll — which lands, not by coincidence, right where flat 1–2% staking already is. For most people, flat staking gets nearly all of Kelly's benefit with none of its fragility.

Measuring your results: ROI and CLV

ROI (return on investment, sometimes called yield) is profit divided by the total amount staked:

ROI = total profit / total staked × 100

A hundred one-unit bets at average odds of 1.80 that win 59 times return 59 × 0.80 − 41 = +6.2 units, an ROI of +6.2%. In football betting, a sustained ROI of a few per cent is very good; claims of 20–30% over a long run usually mean a small sample, or selective reporting.

ROI needs a lot of bets before it means anything. Closing line value measures something faster: whether the prices you take are better than the market's final price. It gives a reading on every bet before the match is played, which makes it the better early warning in both directions — see why CLV is the only metric that matters.

Keep a record

Log every bet: date, match, market, your odds, stake in units, result and profit. Kepler's portfolio page does this automatically for every signal you receive, one unit at the published price, and charts it against the model's own record. If you bet at different prices from ours, keep your own log too: the gap between the two is the part you control.

Withdrawals and bankroll rules

• Withdraw on a schedule, not a feeling. For example, take out anything above your starting bankroll at the end of each month, then recalculate your unit from what remains.
• Never top up to chase. If the bankroll falls, the unit falls with it. That is the plan working, not failing.
• Keep betting money separate. A bankroll mixed with everyday money makes it impossible to know how you are really doing.

Mistakes that break a good edge

• Chasing losses with bigger stakes after a bad day.
• Raising stakes after a winning run, then giving it back on the inevitable losing one.
• Combining picks into accumulators. One miss sinks the whole ticket — see singles vs accumulators.
• Judging the plan on a week. Anything under a few hundred bets is mostly noise.

Frequently asked questions

Betting involves risk. Past results do not guarantee future returns, and no staking plan turns a losing strategy into a winning one — it only decides how fast you find out. Only bet what you can afford to lose. 18+.