Volatility, Bankroll and Why Size Depends on Variance Not Edge

What Variance Actually Does to a Staking Plan

Almost every piece of staking advice assumes all bets of equal value deserve equal money. Two selections at 2.10 with a 4% edge get identical stakes, on the reasoning that the edge is the edge. It is not, because equal edge does not mean equal risk. One resolves half the time within a narrow band; the other swings between a small win and a total loss with a long tail of rare disasters. Same expectation, completely different distribution.

That gap is why a bettor with a real edge can still go broke, and why raising your stake after a winning run feels so natural and is so reliably wrong. Sizing by confidence means sizing by the result you happened to produce, which is a random variable. Sizing by variance asks, before the bet lands, how far this particular bet is likely to stray.

Our bankroll management guide sets the 1% to 2% flat baseline. This article is the adjustment layer above it: how much a given bet deserves relative to your average, based on its odds, its market type and its overlap with what you already hold open.

Expected Deviation by Odds

Assume a 3% edge and a 2-unit stake, then look at how far individual outcomes stray from that average.

What the Table Actually Shows

Read the last column. The spread on a single bet nearly triples between 1.25 and 25.00, from about 1.4 units to 3.8 units of typical deviation, on an identical stake. That is the whole argument in one figure. A bettor staking 2 units on a 25.00 correct score and 2 units on a 1.25 handicap holds two very different objects while claiming to run one plan.

The longshot bet will lose 25 times in a row at some point in any serious sample, because the probability of that is not small. At 4.1% per bet, twenty-five consecutive losses has a probability near one in 250 sequences, which across a year of several thousand such bets is close to certain. When it arrives, the bettor has lost 50 units of a 100-unit bankroll. The average bet lost nothing unusual; the plan simply turned a 4% hit rate into a 50% bankroll exposure, twenty-five times over.

The correction is mechanical. Halve the stake as odds lengthen: base stake up to about 2.50, then reduce progressively so a 6.00 bet takes roughly 70% of base and a 12.00 takes around 50%. You give up some expected return on longshots in exchange for a drawdown profile you can survive.

Sizing by Game Variance, Not by Confidence

Odds are only one variance input. The nature of the fixture matters as much, because some games have near-fixed outcomes and some are structurally open.

A Serie A meeting between two mid-table sides with two fixtures left and neither able to play for a dead rubber carries a narrow distribution: the likely result sits in a small range and the price barely moves. A League A fixture in February carries a wide one, because form, travel, squad rotation and motivation can shift the price 40% in a day and the eventual margin can be anything from four goals to nil.

That produces two different approaches to the same fixture. Pre-match analysis on a wide-variance game means larger stake swings and shorter holding periods. In-play betting on the same game means the price has absorbed most of the information and the edge has to come from timing rather than selection. Our expected goals guide covers the pre-match side; the staking point is that the stake should shrink when the price is doing the work.

The common error runs the other way: staking biggest on the fixture that looks least certain. A 1.20 favourite in a dead rubber is nearly certain and worth almost nothing at 1.20. A 4.50 away win at a side with nothing to play for is uncertain, but if it is genuinely worth 4.50 the price already says so. Confidence in the outcome and confidence in the value are different claims, and only one justifies a stake.

Rules for a Variance-Adjusted Staking Plan

Six rules, applied in order. They are deliberately mechanical, because the point is to replace a feeling with a procedure.

• Calibrate the base stake on your average odds, not your best ones. If most volume sits between 1.80 and 3.00, calibrate to 2.50. Calibrating to the shortest price you ever take is how a plan ends up overexposed to longshots.
• Apply an odds band multiplier. 2.50 and under: 1.0x. 2.51 to 4.00: 0.85x. 4.01 to 7.00: 0.7x. 7.01 to 15.00: 0.55x. Above 15.00: 0.4x. Write the bands down before you bet.
• Halve the stake on any bet correlated with an open position. Same match, same team, same round. Two bets that both fail when one striker has a quiet night are one bet.
• Cap exposure by match, not just by percentage. Eleven 1% correct score selections on one fixture is an 11% exposure to a single match, which no percentage rule stops.
• Treat accumulators as longshots. A five-leg at 4.00 has a 4.1% hit rate and belongs at 0.4x base, not at five separate 1% wagers.
• Recalibrate quarterly. If your average odds drift longer, the base stake falls with them. Most staking plans are set once and then silently outgrown.

Stake Multipliers by Market Type

The same idea applied to markets bettors actually use, with a base stake of 2 units calibrated to a 2.50 average price.

Why Favourites Are Not the Safe Option They Feel Like

The emotional pull of a short price is strong and it points the wrong way. A 1.30 bet feels safe because it usually wins, and in the sense that matters it is the least safe thing in the account: its expected return is the lowest and you will place far more of them chasing the same money.

Work it in volume. To expect 100 units of profit at a 4% edge on 1.30 bets you need roughly 2,200 settled bets. The same 100 units at 3.00 needs about 840. The first strategy produces 2.6 times as many results, which means 2.6 times as many chances to lose a long run, and a higher variance on the total despite each bet feeling safer.

There is a second effect that catches experienced bettors. A high hit rate breeds overconfidence. Someone hitting 78% on 1.30 prices concludes they have a system, raises the stake, and then discovers that a hit rate is not an edge. The two are close to independent: a 1.30 bettor with no edge still wins about 77% of the time and still loses money over a season.

If your record is dominated by short prices, the useful question is not whether you are winning but what your closing line record looks like. Our closing line guide covers the measurement, and it is the only number that separates a high hit rate from a real edge.

The Drawdown You Should Plan For

Before you stake anything, decide what you are willing to watch happen. Simulate 500 bets at your real average odds and real edge, and note the worst point-to-point decline rather than the worst month. A 3% edge at 2.50 average odds on 2% flat stakes typically produces a maximum drawdown in the region of 15% to 20% over 500 bets, and an entirely normal run of twelve losers at 2 units removes 24% of a 100-unit bankroll in an afternoon. A plan that survives that is workable. A plan that requires you not to see it is fiction.

Accumulator Variance Is Not What It Looks Like

An accumulator looks diversified and behaves like a single longshot. Five legs at 2.00 each is a 20.00 ticket, which is a 5% hit rate before margin, so it belongs in the same band as a 20.00 correct score bet. Most bettors stake it as though it were five separate 1% wagers, an exposure ten times larger than their stated risk.

Correlation makes it worse. Five legs from the same Saturday slate are not independent: a manager departure affects a whole round, weather affects a region, a red card changes the shape of the coupon. Your effective number of independent bets is closer to two than five.

Three workable responses. Stake per leg and total the multiplier, which is honest but expensive. Take singles on the legs you have actually researched and use an accumulator only where the price is generous relative to the single. Or accept the variance deliberately at 0.3x to 0.4x base, treating accumulators as entertainment purchases outside the staking plan. Our sports trading guide covers expressing a multi-leg view without the accumulator margin, and the arbitrage guide covers reducing outcome variance to zero at a known cost.

Longshots, Ruin Maths and the Square Root Rule

A 4% hit rate creates a structural problem that flat staking does not solve: in a low-hit-rate series the losing run grows without bound in relative terms while the bankroll shrinks, and that combination is how ruin arrives even when the expectation is positive. The stake must therefore fall roughly with the square root of the odds. Doubling the price should cut the stake to about 70% of its previous size, which is what the bands approximate. A bettor who stakes 2 units on everything has not adopted a staking plan; they have adopted a rule that increases exposure as variance rises.

Putting It Together: A Working System

A complete plan applying everything above, written as rules rather than intentions.

Bankroll 500 units, recalculated monthly. Betting frequency is around 40 bets a week, so the base percentage is 1.5% rather than 2%, since more bets means more variance exposure per month. That is 7.5 units at full bankroll.

Base stake 2 units. Calibrated to an average price of 2.50 across your last 200 bets. Apply the odds bands: 2.00 units up to 2.50, 1.70 units from 2.51 to 4.00, 1.40 units from 4.01 to 7.00, 1.10 units from 7.01 to 15.00, 0.80 units above 15.00.

Correlation rule. Total stake on any single match, including every leg, is capped at 4 units. If the card already holds three units on that match, the correct score takes the remaining unit or nothing.

Edge rule. Only bet where you wrote an estimated probability in advance. If the number is not written down it is not an estimate, it is a preference.

Review. Monthly, record the stake-weighted average odds rather than the raw average, because the stakes are what carry the risk. Quarterly, simulate 500 bets at the actual distribution and check the maximum drawdown is inside 25%.

None of this improves your edge by a decimal place. What it does is make the shape of your results predictable, which is the only reason the edge gets to compound instead of being wiped out by a sequence that was always going to arrive. Our low-stake guide covers running this on a small bankroll, and the journal method is what makes the quarterly review possible.

The Honest Summary

Stake size should answer two questions: how much edge do I have, and how far can this result travel from my expectation? Betting guides rarely separate them, which is why most staking plans are built entirely on the first. Adding the second is a small amount of arithmetic for a materially better chance of still betting in six months.

Rating Breakdown

Frequently Asked Questions