IL MARGINEBack to homeCalculators and staking tools
Compare fair prices, test staking assumptions and explore the range of possible returns. Choose a calculator below, or open the football tool for double chance and draw no bet. Calculations run in your browser. Stake and probability inputs are not saved.
Flat stake, published record
Pick a stake and the tool applies our settled strike rate and return on turnover to that many bets. The curve is a simulation, not a replay: it re-runs the same edge four hundred times so you can see the range of results one edge produces, rather than the single ordering that happened to occur. This simplified model assumes independent bets at one representative winning price. It combines stake-weighted ROI with win rate; it is not a reconstruction of flat-stake historical returns or proof of a future edge.
Range of outcomes
Move the pointer across the chart to read any point. The dashed line is the arithmetic average, which is the only thing a single flat projection ever shows you. The bands show simulated outcomes, not confidence intervals for the true edge. Paths stop when the remaining bankroll cannot fund the next full stake; the dashed reference assumes unlimited funding.
House rules
Start from the bank you would genuinely replace if it went, not the balance in one account.
Price the bet first, then size it. A stake is a function of edge and uncertainty, never of confidence or of the last result.
Cut the Kelly fraction until the worst one in twenty drawdown is a number you would keep betting through.
Remove the margin from a reference market before comparing anything to your own number.
State the removal method alongside the fair price, because the methods disagree where it matters.
Record the price you took and the price at the off. Over a season that record is worth more than the profit column.
The point of fractional staking
Full Kelly is the stake that maximises long-run growth when your probability is exactly right. For small edges, half Kelly retains approximately 75% of optimal growth and quarter Kelly approximately 44%. Smaller stakes reduce volatility, but neither profits nor manageable drawdowns are guaranteed.
The asymmetry matters more once you accept that no probability estimate is exact. Size too aggressively and expected log growth can become negative even when individual bets have positive expected value. The point where that happens depends on the inputs. That is why we run a tenth of Kelly on player props and a quarter on tennis, and why the Kelly tab lets you settle bets at a probability lower than the one you staked at.
Questions
Returns applies the strike rate and return on turnover from our settled record to a stake you choose, and simulates the range of results that edge produces. Kelly sizes a stake from your bankroll, the price and your probability, then shows what each fraction costs in drawdown. Fair odds removes a bookmaker's margin from a set of prices using three published methods. Closing line compares the price you took with a de-vigged closing price. All four are arithmetic on numbers you supply. None of them forecasts a result.
Because a line is not what happens. A record of 2,000 bets at a positive return is one ordering of wins and losses out of an enormous number of possible orderings, and the others include long losing runs that would have tested any bankroll. Simulating the same edge four hundred times shows the spread you would need to fund, and the deepest drawdown figure is usually more useful than the profit figure.
Published prices move. If a selection is posted at 2.10 and you take 1.95 an hour later, you have kept the position and given away most of the margin. Restricted accounts, one-bookmaker shopping and late entry all do the same thing. Setting capture below full is the realistic case for most people, but the appropriate setting depends on the prices actually obtained.
The fractions are scenarios to compare, not universal stakes for a sport. Smaller fractions reduce exposure but cannot fix a wrong probability. Consider estimation error, odds, existing bets and the losses you could fund. Move the estimate-error slider to see how a probability overestimate changes the simulated outcomes.
Proportional is the quickest and the crudest. Shin uses a model of informed betting; in two-way markets it matches additive removal. Odds ratio uses a different adjustment. No method is universally best or guaranteed to sit between the others. On a balanced market they agree to within a few thousandths. On a heavy favourite they separate by more than most claimed edges, so always state which method produced a fair price.
Positive fair-close CLV means the taken price was above a margin-adjusted closing benchmark. Repeating that against a credible market is useful evidence about pricing, but it does not guarantee profit. Check matching settlement rules, capture times and coverage alongside actual results.
The Returns tab keeps the last available settled record if a refresh fails. If no current record has loaded, it uses the historical baseline and labels it clearly. The Kelly, Fair odds and Closing line tools use only the numbers you enter, so they work regardless.
Responsible gambling. These tools describe arithmetic, not outcomes. Past performance does not guarantee future results, and only money you can afford to lose should ever reach a betting account. BeGambleAware and GamCare are there if betting has stopped being a choice.