Finding an edge is the easy half. Surviving it is the job.
Almost all betting advice stops at the pick. The harder question is what to risk on it, because a genuine advantage in football, basketball, or baseball is small enough that an ordinary losing run can empty a bankroll long before the advantage ever appears in the results. Bankroll management is the set of sizing decisions you make in advance, on paper, so that the arithmetic gets the run length it needs to work. It is the least exciting page in this series and the one that decides whether the rest of it matters.
A bankroll, and the unit that comes out of it
A bankroll is money set aside for betting and nothing else, sized so that losing all of it changes nothing important. Every clause in that sentence does work. It is not the balance showing in a sportsbook account, because money you have not deposited yet still belongs to the bankroll and money earmarked for rent never did. Fixing the figure is the first act of bankroll management, and it happens away from any board, before there is a game to have an opinion about.
A unit is a fixed share of that bankroll, commonly 1 percent. On a bankroll of 2,000 a unit is 20. The point of the convention is comparability. Two people who each finished an NFL season up 14 units had a comparable season whether their unit was 5 dollars or 500, while being told that someone is up 700 dollars tells you nothing at all until you also know the bankroll it came out of.
One detail of the convention catches people out. A unit describes what you risk, not what you stand to collect. A one-unit wager at -110 puts a full unit at risk to win 0.909 of one, because the decimal form of that price is 1.909 and the profit is only the part sitting above the stake, as reading betting odds covers in full. So a losing wager costs a clean unit while a winning one profits slightly less than a clean unit, which is why an even record in wins and losses is a losing record in units.
What the swing does to a bankroll on the way through
Take a bettor with a real 55 percent win rate on NBA sides, staking one unit flat at -110 every time. Break-even at that price is 52.38 percent, as the juice explainer works out, so the edge is 2.62 percentage points. In units that is 0.55 multiplied by the 0.909 a winner pays, less the 0.45 of the time a full unit is lost, which comes to 0.05 units per wager. The standard deviation of that same single wager is 0.95 units, nineteen times the size of what it earns on average.
Stretch that over a 500-wager winter and the edge accumulates to plus 25 units against a spread of about 21, which is why roughly one season in eight leaves a genuine 55 percent bettor behind. Losing runs are a separate matter of their own: at a more ordinary 53 percent rate any given six-wager stretch comes up all losses about 1.1 percent of the time, which is roughly once in every hundred starting points, so a 500-wager winter should expect several of them. Why winning picks can be bad bets works both of those figures through. For sizing purposes, though, neither one is the number that matters. Where the season finishes decides how you feel in April. How deep it digs on the way decides whether there is still a bankroll in January.
So take the same season in five stretches of a hundred wagers. Each stretch expects 0.05 multiplied by 100, or 5 units of profit, while its spread is 0.95 multiplied by the square root of 100, which is 9.5 units. A hundred wagers landing one standard deviation below expectation therefore finish at minus 4.5 units, and a hundred landing two below finish at minus 14 units, with five chances in the season for that to be the stretch you draw. Now price it against the bankroll. At a 1 percent unit those 14 units are 14 percent of it, unpleasant and survivable. At 3 percent the identical run of ordinary bad luck takes 42 percent, and at 5 percent it takes 70 percent, which is the point where a real edge no longer has the money left to prove itself. The case for a small unit is not temperament. The bad stretch is the same fourteen units whatever a unit is worth, and the only thing the percentage decides is how much of the bankroll goes with it.
What gets to decide the size of a bet
An expensive habit, and a common one, is raising the stake because a play feels stronger. The pattern is always the same: five units go on a starting pitcher's strikeout prop that got called a lock, one unit goes on everything else, and no record anywhere shows whether the plays given that label have historically won at a rate that justified five times the risk. Confidence is being spent as if it were a measured quantity when it has never once been measured. A play can also be a fine opinion at the wrong price, and the gap between your own estimate and the number on the board is what positive expected value measures.
The arithmetic of that habit is unforgiving. Since a one-unit winner at -110 returns 0.909 units of profit, a single five-unit loser needs 5 divided by 0.909, or five and a half ordinary winners, just to get back to level. Do that four or five times across a baseball season on the strength of a feeling and your graded record can finish green while the bankroll finishes red, because unequal stakes quietly break the link between win rate and profit.
The staking method itself involves a real trade-off, and it is worth stating honestly instead of picking a winner. A percentage of the current bankroll recalculates the unit from what is actually there, so it can never mathematically reach zero, since each wager risks a fraction of whatever remains. It pays for that with slow recovery and a moving yardstick: after a 20 percent drawdown every unit is 20 percent smaller, and it takes a 25 percent gain on the reduced bankroll to return to where you started. A fixed unit set once from the opening bankroll keeps your whole season in one currency, so 14 units in April means what 14 units meant in October, and it accepts in exchange that the bankroll can genuinely hit zero. One buys survival, the other buys a readable record, and knowing which failure you would rather have is more useful than being told which is correct.
Four decisions, all of them made in advance
Decide the bankroll first, keep it somewhere separate from spending money, and pick a figure whose total loss would change nothing you care about. Then set the unit as a share of it. 1 percent is a reasonable default and 3 percent is aggressive rather than merely bold, for a reason that has nothing to do with nerve. The edge you are sizing against is an estimate, not a measured quantity, and the two-standard-deviation stretch above costs 14 percent of the bankroll at a 1 percent unit and 42 percent at a 3 percent one. The aggressive version takes close to half the pool during exactly the run of games that would have told you the estimate was wrong. Notice too that the unit size does not change the 5 percent return on money staked at all, because tripling the unit triples the profit and the amount staked together and the ratio between them never moves. All it changes is how violent the ride is in both directions.
Then hold the number still through both directions of a streak. If you chose the percentage method, its recalculation is the rule and everything here applies to changes made on top of it. Raising the stake after four winners and raising it after four losers are the same mistake in different clothing, because both let the last result choose the next stake. What separates a rule from a reaction is that the rule existed first: a percentage unit moves continuously because you decided in advance that it would, while a fixed unit should move only on a date you set, at the turn of a season rather than in the hour after a bad Sunday.
Last, write down three things and keep them where you will actually see them: the bankroll figure, the unit as both a percentage and a cash amount, and the date you set them. The date is the part everyone skips and the part that does the work, because it is the only way to look back at a season and tell a scheduled recalculation apart from a reaction to a bad week. Tracking bets and ROI covers how to keep the results log so it answers questions later, and the EV calculator will price a single wager in the same units you are staking in.
A staking plan is not a safety net
Every rule on this page governs the size of a loss, and not one of them changes whether the wager was any good in the first place. A disciplined 1 percent unit applied to a losing method simply loses the bankroll on a slower and tidier schedule. No staking rule turns money you need into money you can risk, and the moment a bankroll is being topped up from somewhere it should not be, sizing has stopped being the question worth asking. Our responsible gambling page sets out the deposit limits, the self-exclusion routes, and the helpline numbers that exist for exactly the moment a bankroll stops being a fixed pool, and it is worth reading while you are setting the number rather than after you have run through it.
Bankroll management, answered
What is bankroll management in sports betting?
Bankroll management is the set of rules that decide how much you risk on each wager, chosen in advance of any particular game. It exists because a real edge is small while ordinary losing runs are large, so the size of a bet has to be set for the long run rather than for the strength of one opinion. It controls how long you last, and lasting is what gives an advantage the time it needs to show up.
What is a unit in sports betting?
A unit is a fixed share of a bankroll used as the standard size of one wager, commonly 1 percent of it. On a 2,000 dollar bankroll a unit would be 20 dollars. Recording results in units rather than dollars is what makes two records comparable, because 14 units means the same thing at any stake level while 700 dollars means nothing until you also know the bankroll behind it.
How much should I bet per game?
One unit on almost every game is the usual answer, and its value is that it keeps your record measuring your picks rather than measuring how confident you happened to feel on the day. A flat stake also removes a decision that is easy to get wrong, which is how much extra to risk on the play you like most. If you vary the size at all, vary it inside a narrow band you defined before the slate started.
What percentage of my bankroll should I bet on one wager?
One percent of the bankroll per wager is a reasonable default and three percent is aggressive, because your edge is an estimate and a wrong estimate at three percent empties the pool before the error is detectable. The percentage does not change your return on the money staked at all, it changes the size of the swings, so a larger unit magnifies a losing run in the same proportion as a winning one. Those two directions are not symmetrical, though, because only the losing one has a floor, and a bankroll that reaches it is no longer there to bet the recovery.
What is the difference between flat staking and percentage staking?
Flat staking fixes the unit once from the starting bankroll and keeps it there, while percentage staking recalculates the unit from the current balance before each wager. Percentage staking cannot mathematically reach zero, because a fixed share of a shrinking balance is always something rather than nothing, but it recovers slowly because the unit shrinks along with the bankroll. Flat staking keeps a whole season of results in one consistent currency, and the price it pays is that the bankroll really can reach zero.
Should I bet more on plays I am more confident in?
Betting more on a play you like more is defensible only if that confidence has been measured rather than felt. At -110 a single five-unit loss takes five and a half ordinary one-unit winners to repair, so sizing up on a label is an expensive way to express an opinion. Unless you hold a record showing that your highest-confidence plays have won at a rate that justifies the extra risk, the label is a feeling and the stake is a guess.
What should I do after a losing run?
After a losing run the correct action is to leave the stake exactly where it is, which is the hard part. A run of six straight losses turns up roughly once in every hundred starting points at a 53 percent win rate, so across a few hundred wagers it is an expected event rather than evidence that anything has broken. Chasing raises the stake at the exact moment the bankroll can least afford it, and cutting it out of fear, as opposed to by a schedule you set in advance, means a real edge earns less on the recovery.
How big should my starting bankroll be?
There is no formula for this, and anyone offering one is selling something. The honest test is that the whole amount could be lost without changing anything about your month, which makes the right figure a personal answer rather than a mathematical one. What matters more than the size is that it is a fixed pool set aside in advance, because a bankroll that gets topped up whenever it runs low is not a bankroll.
Decide the risk against a real number
Our boards put an independent projection next to the market price on every row, so the size of the disagreement is something you can look at before you decide how much of the bankroll it is worth.