Bankroll Calculator

Turn your bankroll and unit size into a flat stake, and see what a long losing run leaves of it.

This shows what a staking plan does to a bankroll. It cannot create an edge, and at a win rate of 50% or below on an even-money bet there is no plan here that makes the run profitable.

Recommended stake

$40.00

2% · 50 units in the bankroll

Bankroll$2,000One bet$40.00Units in bankroll50

1. Set the bankroll

Bankroll$2,000

2. Lose a streak, on both plans

Losing streak10 losses in a row

Flat staking$1,600 left (80%)

Next bet is still $40, unchanged by the streak.

Proportional staking$1,634 left (82%)

Next bet is $32.68, down from $40. The smaller bet is what bought the extra room.

Flat: bets to bust

Proportional: bets to halve

Flat staking risks the whole bankroll in a fixed number of losses. Proportional staking shrinks each bet as you lose, so it bleeds slower but never fully recovers to the same stake without winning. Shrinking the bet is what buys the extra room, not any protection of the bankroll: the balance keeps heading the same way, in smaller steps.

What 200 bets at this plan actually look like

Your win rate50% at even money

Simulating…

Both plans, every preset stake

The highlighted cell is the reading above. The other seven run the same simulation at every other stake and plan, so the failure each one buys is a figure rather than a sentence.

Share of 200-bet runs that busted, and the typical (median) finish, flat and proportional staking across the four preset stakes, at a $2,000 bankroll and 50% win rate. All cells at 600 trials.
Plan	1% stake	2% stake	3% stake	5% stake
Flat	Simulating	Simulating	Simulating	Simulating
Proportional	Simulating	Simulating	Simulating	Simulating

Neither plan is protection at or below break-even: you have no edge, so no staking plan here makes the run profitable. The grid above shows what each plan does with that fact, not a way to escape it.

Beat the downswing

$1000 bankroll, a 47% coin at even money. Double it to $2,000 before you fall below $50. Pick a stake each bet, bigger is faster and closer to the cliff.

Bankroll

$1,000

0 of 60 bets

Share this scenario

A link to this page opens the calculator on the same bankroll, stake size, win rate and plan, with the same notes and limits shown here. It carries those inputs and nothing else: nothing is stored, and nothing in it identifies you.

Written and reviewed by Alexander Korsager · last reviewed September 25, 2026 · corrections

The idea

A bankroll is the money you have set aside for betting and can afford to lose. Staking a fixed percentage of it on each bet keeps you in the game through losing runs: bets shrink automatically when you are down and grow as you recover.

This calculator turns a bankroll figure and a risk level into a flat stake, then simulates what 200 bets at that stake actually do to the balance.The two staking plans it offers fail in different ways, and the difference is the thing worth knowing before picking a percentage: flat staking leaves the typical run level and raises the chance of running out, while proportional staking almost never runs out and grinds the typical run down instead.
That matters because the usual advice, stake 1% to 2%, is a risk-of-ruin argument, and risk of ruin is only the flat-staking half of what this page models.The table below the simulation sweeps both plans across all four preset stakes so you can see which failure each percentage is buying.

FAQ

Questions players ask

What percentage of my bankroll should I bet?

The usual answer is 1% to 2%, and it is a risk-of-ruin argument, which describes flat staking. The sweep on this page runs both plans across all four preset stakes at your own bankroll and win rate. At the default fair coin (no edge), flat staking's chance of running out climbs as the stake grows while its typical finish barely moves, and proportional staking's typical finish falls while its chance of running out barely moves. With a real edge both figures can move together instead, which is exactly what the sweep is for: read it at your own inputs rather than assuming the fair-coin shape. The percentage is a choice between failures at no edge, not a single safe number, and the table shows which one each stake is buying at the win rate you set.

Why does a bigger stake cost so much more under proportional staking?

Because the cost scales with the square of the stake, not with the stake. A win and a loss of the same percentage do not cancel: a 5% gain and a 5% loss leave 99.75% of the bankroll, and the shortfall is the stake fraction squared. Over 200 bets at the default fair coin that happens up to 100 times, so a stake five times bigger costs far more than five times the money. The sweep table above shows the typical finish at every preset stake side by side, rather than one pair of figures.

Flat staking or percentage staking?

Percentage staking recalculates the stake as the bankroll changes, so the bet shrinks on the way down and grows on the way up. It compounds in both directions, and with no edge that means downwards: the typical run finishes below where it started at every stake this tool offers. Flat staking keeps the bet fixed, which holds the typical run level and puts the risk into running out instead. Neither is protection, and the simulation runs both on the same inputs so you can compare them rather than pick on reputation.

How is this different from the Kelly calculator?

This uses a fixed risk level you choose. The Kelly tool derives the stake from your actual edge and the odds, use it when you can estimate your edge.

Methodology

How this is calculated

The stake, the unit count and the losing-streak figures are exact arithmetic: a stake is the chosen percentage of the bankroll, and the streak panel applies that many consecutive losses with nothing in between.

The distribution is a Monte Carlo simulation instead: 3,000 runs of 200 even-money bets at the staking plan and win rate you set, recording what share finished ahead, where the bankroll ended, and the deepest fall from a peak along the way. One figure in that panel is not really random.

Under proportional staking the finishing bankroll depends only on how many of the 200 bets were won and not on the order they arrived in, so the middle run is the run with the middle win count, and the typical finish the panel prints is the starting bankroll multiplied by (1 plus the stake fraction) once for each winning bet and by (1 minus the stake fraction) once for each losing one.

You can reproduce it on a pocket calculator rather than taking the simulation's word for it, using the win count your own win rate implies.

At the default 50% that count is 100 wins against 100 losses, the two halves pair off, and the figure collapses to the starting bankroll times (1 minus the stake fraction squared), 100 times over.

That collapse is the fair-coin case and nothing more general: at a 55% win rate the middle run is 110 wins against 90 losses, so a 5% stake on $2,000 finishes at $4,236, where the squared form would have told you $1,557.

Both forms also assume the middle run never falls below the point where its next bet is unplaceable, at any stage rather than only at the end, since a run that stops early no longer depends on its win count alone. At a 7.5% stake that assumption holds down to a 44.5% win rate and fails at 44%.

Common misreadings4

A staking plan is not risk management on its own. Sizing bets smaller extends how long a losing bankroll lasts, it does not change the direction it is heading.
Proportional staking never reaching zero mathematically is not protection. It reaches a balance too small to place the plan's next bet, which is the same outcome in practice.
A clean bust record is not a verdict on the stake. None of 3,000 proportional runs at any of the four preset stakes fell too low to place the next bet, at 1% and at 5% alike, while flat staking went from 0.0% to 15.6% across the same four. Reading the first column as safety inverts what the second one costs: at 5% the typical proportional run still finished $443 down on a bankroll that was never in visible danger. The clean record also belongs to the presets rather than to the plan. At the top of the custom range, 10% a bet, 26.1% of proportional runs ran out anyway and the typical one finished at $732.
The average ending bankroll is not the typical one. Under proportional staking they come apart, because a win and a loss of the same percentage do not cancel. The result panel prints both, with the share of simulated runs that finished below the average, so you can see how far apart they are on your own inputs rather than taking this on trust.

Assumptions5

Every bet is even money and independent, won at the win rate you set. The default of 50% is a fair coin, which is a losing proposition against any real price and is the honest starting point for a player with no demonstrated edge.
Flat staking bets the same amount every time, computed from the starting bankroll. Proportional staking recomputes the stake from the current bankroll before each bet.
A run is called busted when the bankroll falls below one starting stake, since the next bet the plan calls for can no longer be placed. Without that floor, proportional staking would never technically reach zero.
3,000 simulated runs of 200 bets, run in your browser from a fixed seed, so the same inputs always produce the same picture.
Drawdown is measured against the highest point a run reached, not against where it started.

Limits4

200 bets is a stated stretch of betting, not a career. A longer run at a losing win rate ends worse, not better.
Even-money bets only. A staking plan applied to longer odds has a different shape, because the win rate and the payout both move.
The streak panel and the simulation answer different questions. The streak figures are the worst ordering of a fixed number of losses; the simulation is the whole space of orderings.
The deepest-fall band is not where the two plans differ. At a 5% stake on the default coin the typical run falls 57.1% from its peak under flat staking and 58.3% under proportional, which is as close as the two get anywhere on this page. Whatever room proportional staking buys shows up in dollars left at the end, not in a shallower ride to get there.

Evidence

The simulation is covered in simulate.test.ts, which checks that flat staking at a fair coin holds the mean bankroll flat, that proportional staking drags the median below the start and drags harder at a larger stake fraction, and that flat staking busts more often than proportional at the same fraction.

It also pins the gap between the average and the typical run in both directions: clear under proportional staking, absent under flat staking on the same coin.

It also pins the claims this page leads on: that the simulated proportional median matches the win-count form above to the cent across the win-rate slider as well as at every preset stake, that the squared shortcut is true only at the fair-coin default and how far off it goes elsewhere, that the win-count form itself stops holding once the middle run starts breaking off early, and the split in how the two plans fail, with flat staking's bust rate climbing while its typical finish holds and proportional staking's typical finish falling while its bust rate does not move.

A passing test means the code matches the model, not that the model fits every real betting pattern.

Gambling can cause serious harm. The National Problem Gambling Helpline is free, confidential and open 24/7, call or text 1-800-MY-RESET (1-800-697-3738), or chat at ncpgambling.org/chat.

Read up

Go deeper

Related tools