Staking Plans: Flat, Percentage and Kelly
A staking plan decides how much of your bankroll goes on each bet. It cannot make a losing set of tips win, but it decides whether a winning set of tips survives its losing runs. This guide compares the three plans worth knowing and runs each of them through the deepest drawdown in this site's own record.
Flat stakes
The same stake on every bet, usually 1% or 2% of the starting bankroll, called a unit. It is the simplest plan and the one every published record on this site uses. Its great virtue is that the record then describes the tips and nothing else: a return of +4.0% on stakes means exactly that, and a drawdown of a given number of units is exactly that many units of money. The weakness is that the stake does not shrink in a losing run, so a bankroll that is too small can be emptied.
Percentage of bank
A fixed share of the current bankroll, recalculated after every bet. Stakes fall during a losing run and rise during a winning one, so the bank can shrink indefinitely but never reach zero. The cost is on the way back: after a 30% fall the stakes are 30% smaller, so it takes more winners to regain the old high than it took losers to leave it. Over a long, mostly sideways record this plan quietly underperforms flat stakes.
Kelly
The Kelly criterion sets the stake from the edge and the odds: stake the fraction (true probability × odds − 1) ÷ (odds − 1) of the bank. At odds of 2.00 and a true chance of 55% that is 10% of the bank, which is why nobody uses full Kelly; a quarter or a tenth of it is normal. Kelly grows a bank faster than any other plan if the probability estimate is right, and punishes an over-estimate harder than any other plan. A follower of tips does not know the true probability of each selection, only the record's average, which makes Kelly a poor fit for following and a better fit for pricing.
Through the record's deepest drawdown
The worst stretch in the record so far fell −109.6 units from a high over 1,403 tips, then took 1,021 tips to regain the high. The table starts each plan with a 200-unit bank and runs it through that stretch.
| Plan | Stake | Through the drawdown | Recovery | Character |
|---|---|---|---|---|
| Flat stakes | 1 unit on every tip (0.5% of a 200-unit bank) | 200 → 90 units | Back to the previous high after 1,021 tips | Linear. The drawdown in units is exactly the drawdown in money. |
| Percentage of bank | 1% of the current bank on every tip | 200 → 66 units | Slower on the way back up, because stakes shrink as the bank shrinks | Cannot reach zero, but a long drawdown is followed by a longer climb. |
| Kelly (fractional) | A quarter of the Kelly fraction, which varies with each price | Depends on the mix of prices; larger stakes on shorter prices | Fastest when the edge estimate is right, worst when it is wrong | Needs a reliable estimate of the edge per bet, which a follower does not have. |
The lesson is in the first column. A 200-unit bank at flat stakes came through the drawdown with 90 units; a 55-unit bank at the same stake would have been emptied, and so would a 100-unit bank. Pick the unit so that the bankroll is comfortably larger than the deepest drawdown the record has shown, and assume the next one will be deeper.
Plans to avoid
Anything that raises the stake after a loss. Martingale (double after each loss) and its gentler cousins promise to recover every losing run with one winner, and they do, until a run arrives that the bankroll cannot double through. At the record's strike rate of 55.4% the losing runs guide shows how often runs of five and eight have occurred. A doubling plan through eight losses needs 255 units to win back one.
Every settled tip is published with its odds and result, so any plan can be tested against the real sequence using the tips CSV.
Frequently asked questions
What is a unit in betting?
A fixed fraction of your bankroll that every stake is expressed in, so that results can be compared regardless of how much money is involved. One unit is commonly 1% to 2% of the bankroll. A record quoted in units, as this site's is, means every tip was staked at the same size.
Does a staking plan turn a losing record into a winning one?
No. Staking changes how big the swings are and how quickly a bankroll grows or empties; it cannot change the sign of the expected value. Progressive plans such as Martingale, which raise the stake after a loss, only shift when the loss is taken and make it larger.
Why does the record use flat stakes?
Because it is the only staking method under which the figures describe the tips rather than the staking. Any reader can apply their own plan to the same tips, since every tip is published with its odds and result.
More guides
- Which Odds Ranges Win Most Often · Strike rate and return for every price band in the record, against what the odds implied.
- Results by Betting Market · Match winner, totals, handicaps, both teams to score and props: how each market has performed.
- How Long Losing Runs Last · The longest winning and losing runs in the record, the biggest drawdown, and what that means for a bankroll.
- Expected Value in Betting, Explained · What a positive expected value bet is, how implied probability works, and how to read it off a real record.
- Closing Line Value and Why Odds Move · Why the price at kick-off matters, how to measure whether you beat it, and what it can and cannot tell you.
- All guides and monthly results
18+ only. Please gamble responsibly. Past results do not guarantee future outcomes. Figures on this page are recomputed from the public record as tips settle.