Bankroll Management Betting Guide: Staking Plans That Survive a Losing Run

Bankroll management in betting means deciding, before you place a bet, how much money you are prepared to lose in total and what fixed, small share of it each bet will risk. It does not make winning more likely. What it decides is how long your money lasts through a losing run, and losing runs are a certainty: on bets that win half the time, most people who place 500 bets will lose eight in a row at least once. This guide shows how the common staking plans behave when that happens, using a simulation of 100,000 bettors, the published maths on the Kelly criterion and the reasons the Martingale fails.

There is no tip or winning system here. The simulation code and its numbers are our own. The Kelly figures come from a paper by Edward Thorp, the mathematician who applied Kelly’s formula to blackjack and sports betting, and the losing-streak arithmetic is worked out exactly rather than estimated.

Start With the Bankroll, Not the Bet

A staking plan only works if the pot it draws from is fixed. The bankroll should be money you have already accepted losing, set aside for a week, a month or a season, and kept separate from anything you need. If losing all of it would mean a missed bill or a credit card balance, it is too big.

GambleAware puts the first rule plainly: decide your spending limit for the day, week or month before you start, and commit to it. It also suggests writing the limit down or telling someone you trust. That is not a separate safety step from bankroll management. It is the first step of it, because every percentage in this guide is a percentage of that number.

Two habits make the bankroll real rather than notional:

• Keep it in one place. One account, one balance. Topping up from your everyday account mid-month turns a fixed bankroll into an open-ended one.
• Do not refill it after a bad week. If the month’s money is gone, the plan for the month is finished. Reloading is how a staking plan quietly becomes chasing.

Our beginner’s guide to sports betting covers budgeting and record-keeping from the very start. This page goes further into the staking question: once the bankroll exists, how big should each bet be?

Why Losing Runs Are Longer Than They Feel

Most people underestimate how long an unlucky run can be. A string of eight losses feels like evidence that something is broken. On a coin-flip bet, it is ordinary. We calculated the exact chance of seeing at least one losing run of a given length on bets that win 50% of the time:

Losing run of at least…	Across 100 bets	Across 500 bets	Across 1,000 bets
5 losses	81.0%	almost 100%	almost 100%
7 losses	31.8%	86.4%	98.2%
8 losses	17.0%	62.5%	86.1%
10 losses	4.4%	21.5%	38.5%
12 losses	1.1%	5.8%	11.4%
15 losses	0.1%	0.7%	1.5%

A run of five losses is close to certain within a hundred bets. Over a few seasons of weekend betting, which can easily reach 500 or 1,000 bets, a run of eight is more likely than not, and a run of ten is a real possibility.

These figures are for bets that win half the time, such as many Asian handicap and over/under lines priced near evens. Bets at longer odds lose more often, so their losing runs are longer. Wikipedia’s article on the Martingale system gives the casino version: on a single-zero roulette wheel, where an even-money bet loses 19 times in 37, there is about an 84% chance of a six-loss run somewhere in 200 spins, and about an 11% chance of a ten-loss run.

Losing runs do not mean a bettor is unlucky in some special way, and they do not mean a win is “due”. Each bet is independent. The runs appear simply because, over enough bets, unlikely sequences get many chances to happen.

Five Staking Plans, One Losing Run

Here is what a single run of ten straight losses does to a 100-unit bankroll under the staking plans people actually use. The Kelly stake assumes a bettor who believes they have a 55% chance at odds of 1.91, which we explain below.

Staking plan	First stake	Bankroll left after 10 losses	Fall
Flat 1% of the starting bankroll	1 unit	90 units	10%
Percentage: 2% of the current bankroll	2 units	81.7 units	18.3%
Half Kelly on an estimated 55%	2.8 units	75.5 units	24.5%
Full Kelly on an estimated 55%	5.5 units	56.5 units	43.5%
Flat 5% of the starting bankroll	5 units	50 units	50%
Martingale from 1 unit	1 unit	Next bet needed: 1,024 units	1,023 units lost

The percentage plans lose less than flat staking of the same size, because each loss makes the next stake smaller. That is their main strength. The flat 1% plan loses least of all because its stake is the smallest. Size matters far more than the formula behind it.

Flat staking

Every bet is the same amount, set as a share of the bankroll you started with: 1 unit from 100, for example. It is the simplest plan and the easiest to keep to. The downside is that it does not shrink as the bankroll does, so if you start too high, a bad run can empty the account. A flat 1% stake needs a hundred straight losses to wipe out the bankroll. A flat 5% stake needs only twenty.

Percentage staking

Every bet is a fixed share of whatever the bankroll is now. After a loss the stake falls, and after a win it rises. In theory it can never reach zero, because each bet takes only a fraction of what remains. In practice there is a minimum stake, and the bankroll can shrink to the point where betting is pointless. The other catch is recovery: after an 18.3% fall, the remaining 81.7 units need to grow by about 22% just to get back to 100, and the smaller stakes make that slow.

Kelly staking

The Kelly criterion, published by John Kelly of Bell Labs in 1956, sets the stake from your edge. For a bet that loses the whole stake when it loses, the formula is the chance of winning minus the chance of losing divided by the net odds. A bettor who believes a bet at 1.91 wins 55% of the time works it out as 0.55 − 0.45 ÷ 0.91 = about 0.055, so 5.5% of the bankroll. Half Kelly bets half that.

The trouble is the input. If the true chance of that bet is 50%, the same formula gives 0.50 − 0.50 ÷ 0.91, which is negative. The correct Kelly stake on a bet with no edge is nothing at all. Kelly only tells you how much to bet once you know your edge, and an inflated estimate produces an inflated stake. Our value betting guide explains why an edge is so hard to measure, and why a few months of good results cannot show one.

Martingale and other progressions

The Martingale doubles the stake after every loss so that the first win wins back everything plus one unit. After ten losses in a row, a bettor who started at 1 unit has lost 1,023 units and needs to stake 1,024 on the next bet. The idea is old and its flaws are well known. Wikipedia’s article on it notes that the system fails whenever there is a limit on what you can bet or how much money you have, and that in a game where each bet is more likely to lose, larger stakes only increase the average loss. Our roulette odds guide covers the Fibonacci and Labouchère variations, which grow more slowly and fail for the same reason.

What 100,000 Simulated Bettors Show

One losing run is a snapshot. To see how each plan holds up across a longer stretch, we simulated 100,000 bettors placing 500 bets each. Every bet was on a contest with a true 50% chance at odds of 1.91, a typical price for an even contest once the bookmaker’s margin is added, which gives an expected loss of 4.5% of each stake. Every bettor started with 100 units, and the same sequence of wins and losses was used for every plan, so the plans are compared on identical luck. A bettor who could not afford the next stake stopped there. The random seed is fixed so the run can be repeated.

Staking plan	Median bankroll after 500 bets	Ahead after 500 bets	Fell to half the bankroll at some point	Could not place the next stake, or fell below 10 units
Flat 1 unit (1%)	77.5	15.3%	14.3%	0%
2% of current bankroll	58.2	9.8%	47.0%	0%
Half Kelly on an assumed 55%	44.9	8.3%	70.2%	0.9%
Full Kelly on an assumed 55%	12.7	4.9%	94.1%	47.5%
Flat 5 units (5%)	3.8	13.8%	88.5%	72.4%
Martingale from 1 unit	49.9	16.0%	50.3%	88.7% could not cover the next stake

Four things stand out.

No staking plan changed the direction. Every plan’s median bettor finished behind. The flat 1-unit bettor lost 22.5 units at the median, which is exactly 500 bets × 1 unit × 4.5%. A staking plan decides how quickly the margin is collected, not whether it is.

Small stakes bought time, not profit. The flat 1% plan kept every bettor in the game for all 500 bets, and only one in seven ever saw the bankroll halve. At 5% the same luck left almost three in four bettors unable to place a 5-unit bet before the end.

A wrong edge was punished hardest. Full Kelly on an imagined 55% chance was the worst percentage plan by a wide margin: nearly half of those bettors were below a tenth of their bankroll after 500 bets. Halving the stake made a large difference, though half Kelly still did worse than a plain 2% because its stake was bigger.

The Martingale looked best right up until it didn’t. Its 16.0% share of bettors ahead is the highest in the table, because most cycles end with a small win. But 88.7% of Martingale bettors hit a run that left them unable to double again within 500 bets, and the median stake they reached was 64 units, almost two-thirds of the starting bankroll on a single bet.

What Thorp Found About Betting Less Than Kelly

Edward Thorp used the Kelly criterion for blackjack and later for a test of a sports betting system, which he wrote up in a chapter of the Handbook of Asset and Liability Management published by Elsevier in 2006. His conclusions about bet sizing are the most useful part for anyone thinking about staking.

He wrote that “using too large an f* and overbetting is much more severely penalized than using too small an f* and underbetting”, where f* is the Kelly fraction. In his idealised model, betting half the Kelly amount keeps three-quarters of the growth rate while sharply cutting the risk of a deep fall: the chance of ever losing half the starting capital is 1 in 2 at full Kelly, but 1 in 8 at half Kelly.

He also showed what happens when the edge is overestimated. If the true edge is half what you think and you bet full Kelly on your estimate, his model gives zero growth and “increasingly wild oscillations” around the starting capital. Bet 50% more than that and you are ruined. His advice was that, to the extent an estimate is uncertain, you should assume the true edge is smaller and bet less.

Two points are worth keeping in mind. Thorp’s results apply to bettors who had a measurable edge, which in blackjack can be calculated from the rules and in his sports test came from a model he trusted. And even in that best case, his recommendation was to bet less than the formula says. For a bettor without a proven edge, the Kelly stake is zero, and the case for small stakes is stronger still.

Gambler’s Ruin: Why the Bankroll Always Has a Limit

The maths of gambler’s ruin explains why staking plans cannot rescue a negative-edge bet. Wikipedia’s summary of the result is that a gambler with finite money playing a fair game against an opponent with unlimited money will eventually go broke, and that a game with non-positive expected value will bankrupt the player eventually “regardless of their betting system”. A bookmaker is not infinitely rich, but compared with one bettor’s bankroll it might as well be, and its prices carry a margin.

That is why the honest goal of bankroll management is modest. It is a way to make a fixed entertainment budget last, to limit the damage a bad run can do, and to make losses predictable enough that they never come as a shock. A bet calculator helps here in a practical way: entering your stake and odds before you bet shows the exact return and how much a loss costs, which keeps the stake tied to the plan rather than to the mood of the moment.

A Simple Staking Plan You Can Actually Keep To

None of this needs a formula. A plan that survives a losing run usually looks like this:

1. Fix the bankroll for a set period. A month or a season, money you can lose without consequence.
2. Choose one stake size and write it down. Around 1% to 2% of the bankroll is a common choice. If ten losses in a row at that stake would upset you, make it smaller.
3. Keep the stake the same after losses. Never raise it to win back money. GambleAware’s warning is direct: trying to win back your losses often leads to even bigger losses.
4. Keep the stake the same after wins too. A good week is not evidence of skill, and a bigger stake makes the next bad run more expensive.
5. Keep a record. Date, market, odds, stake, result. After a few hundred bets your real return will be clear, and it is usually lower than memory suggests.
6. Stop when the bankroll for the period is gone. GambleAware’s advice is that once you have lost the money you set as your limit, it is time to walk away.

Accumulators and long-odds bets need extra care, because they lose far more often than they win. If you bet on them, treat the stake as its own small bankroll rather than mixing it with the even-money bets above.

When Staking Stops Being the Problem

Bankroll management assumes you can stick to the rules you set. If you find yourself raising stakes after losses, betting with money meant for something else, refilling the bankroll mid-month or hiding how much you bet, the problem is no longer staking. Gambling is for adults aged 18 and over, and for many people the healthiest plan is to stop. Free, confidential help is available, and our responsible gambling resources list support services and self-exclusion options by country.