Bankroll Management: What Our Own Tips Would Have Done to Your Money
How much you stake decides whether a bad run is an annoying month or the end of your bankroll. We replayed every main tip in our public archive in order, with different staking plans and a starting bankroll of 100, to show what really happens.
What a betting bankroll is
Your bankroll is the money you set aside for betting and nothing else: not rent, not savings. You measure every stake as a share of it. If your bankroll is 500 and one unit is 1%, a unit is 5.
The point is not to win more. It is to survive the losing runs that come with every betting record, including ours, without ever having to chase losses or top up from money you need.
Our real record, month by month
Since 11 May 2026, the start of our public tips archive, we have published 328 main tips. At average odds of 1.37, the total was +4.56 units, a return of +1.4%. Month by month:
| Month | Tips | Units | Return | |
|---|---|---|---|---|
| May 2026 | 40 | +4.6 | +11.5% | |
| June 2026 | 57 | −2.4 | −4.2% | |
| July 2026 | 64 | −2.2 | −3.4% | |
| August 2026 | 84 | −2.5 | −3.0% | |
| September 2026 | 64 | +8.6 | +13.5% | |
| October 2026 | 19 | −1.6 | −8.2% |
Source: Daily Sport Pick main tips, 11 May to 12 October 2026, 1 unit per tip, settled at the odds published. May starts on the 11th, October is not complete. Every tip is in our archive.
The longest losing run was 4 tips in a row, and at a flat 1 unit per bet the bankroll fell at most 9.6 units from its peak. A small profit, with real dips on the way: June, July and August all ended slightly negative before September made up for it.
One month can look brilliant (+13.5% in September) and the next disappointing (−8.2% in the first weeks of October). Judge any tipster, including us, on hundreds of bets, never on a month.
Staking tested on our tips
We replayed all 328 main tips in the order they were published, starting from a bankroll of 100. Percentage stakes are calculated from the bankroll at the start of each day:
| Staking plan | Final bankroll | Worst drawdown | Lowest point |
|---|---|---|---|
| Flat 1 unit per bet | 104.6 | −8.7% | 96.7 |
| Flat 2 units per bet | 109.1 | −16.5% | 93.4 |
| Flat 5 units per bet | 122.8 | −35.6% | 83.5 |
| Flat 10 units per bet | 145.6 | −58.0% | 66.9 |
| 1% of bankroll per bet | 104.1 | −9.1% | 96.3 |
| 2% of bankroll per bet | 107.1 | −18.0% | 91.9 |
| 5% of bankroll per bet | 108.9 | −42.1% | 75.4 |
| 10% of bankroll per bet | 87.5 | −72.1% | 44.3 |
Simulation on our published main tips, 11 May to 12 October 2026, starting bankroll 100.
The tips were the same in every row and made a small profit. At 1% per bet the bankroll never fell more than 9%. At 10% per bet it fell 72% from its peak and ended at 87.5: a loss on profitable tips.
Bigger stakes did not just raise the profit; they raised the dips much faster. A flat 10 units ended highest, but only after a fall of 58% from its peak, which few bettors would sit through. At 10% of the bankroll, the losing spells shrank the stakes just before the winning ones, and the bankroll ended below where it started. No staking plan turns a losing record into a winning one, and on a thin edge a big stake can turn a winning record into a loss.
The Kelly criterion, and why it bankrupted our model
The Kelly criterion tells you what share of your bankroll to stake to grow it fastest, given your edge:
A 55% chance at odds of 2.00: (0.55 × 2.00 − 1) / (2.00 − 1) = 0.10, so stake 10% of your bankroll.
Kelly is mathematically optimal if your probabilities are right. We tested what happens when they are not, using the 863 value bets from our value betting test, where our model claimed large edges but the bets returned −9.1%. Based on the model’s probabilities, Kelly recommended an average stake of 30.7% of the bankroll per bet:
| Staking plan | Final bankroll (from 100) | Worst drawdown |
|---|---|---|
| Full Kelly | 0.01 | −100.0% |
| Half Kelly | 0.01 | −100.0% |
| Quarter Kelly | 0.01 | −100.0% |
| Tenth Kelly | 3.69 | −96.9% |
| 1% of bankroll per bet | 43.33 | −61.3% |
| 2% of bankroll per bet | 16.84 | −86.3% |
| Flat 1 unit per bet | 21.66 | −81.1% |
Simulation on our daily value table, 12 May to 12 October 2026, 863 settled bets, model probabilities as input for Kelly.
Full, half and quarter Kelly all went bust. Even a tenth of Kelly lost 96.9%. Kelly does not protect you from a wrong model; it magnifies it.
If you use Kelly at all, use a small fraction of it and probabilities you have checked against real results, not ones that only look good on paper. For most bettors a fixed 1% per bet is safer and easier.
The rules we would follow
- Stake 1 to 2% per bet. In our test, 1% never fell more than 9% from its peak. At 5% the worst fall was 42%, at 10% it was 72%.
- Think in units, not money. Decide your unit once and keep it, so a losing run never tempts you into bigger bets.
- Never chase losses. Doubling up after a loss is the fastest way to turn a bad week into an empty account.
- Keep a full record. Every bet, at the odds you took. Without it you cannot see the slow drift that hurts most.
- Set limits with your bookmaker. Deposit and loss limits are free. Only bet if you are 18 or older, and if betting stops being fun, BeGambleAware offers free, confidential help.
Frequently asked questions
What percentage of my bankroll should I stake per bet?
One to two percent. When we replayed our 328 main tips since 11 May 2026, 1% per bet ended at 104 of a 100 bankroll with a worst fall of 9%. 5% ended at 109 but fell 42% on the way, and 10% ended at 87.5 after a fall of 72%. The tips were the same; only the stake size changed.
Is flat staking or percentage staking better?
Both work at small sizes: 1% of bankroll ended at 104.1, a flat 1 unit at 104.6. At larger sizes they part ways: a flat 10 units ended at 145.6 after a 58% fall, 10% of bankroll at 87.5 after a 72% fall. Percentage staking shrinks your stakes after losses, which also means smaller stakes when the record recovers.
Should I use the Kelly criterion?
Only a small fraction, and only with probabilities you have verified. With our own model probabilities, full, half and quarter Kelly all went bust on 863 bets, and a tenth of Kelly lost 97%, because the model overestimated its edge.
How long can a losing run last?
Longer than most people expect. Our longest run since 11 May 2026 was 4 losing tips in a row, and the worst drawdown at 1 unit per bet was 9.6 units. June, July and August all ended slightly negative before the record turned.
Should I stake more on bets I am more confident about?
Be careful. In our own data, our most confident picks were consistently less accurate than our model claimed. Raising stakes on confidence only helps if that confidence is proven to be right.
Get free tips, every day
Daily AI-backed football tips in your inbox. Free, no spam, unsubscribe anytime.
18+ only. Bet responsibly. Gambling can be addictive.