Bankroll management: how to size your stakes so variance cannot break you

Every losing player in a casino or poker room can name a hand, a spin or a bad beat that "cost" them. Almost none of them can name the real cause: they were playing stakes their bankroll could not support, so an ordinary bad run became a wipe-out. Bankroll management is the set of rules that stops that happening. It does not make a losing game winnable, but it makes a winning game survivable — and it keeps a losing hobby affordable.

Unit sizing: the one number to decide first

A unit is your standard stake, expressed as a fraction of your bankroll. Everything else — buy-ins, session budgets, stop-losses — is built out of units. The right fraction depends on two things: your edge (positive, zero or negative) and how swingy the game is.

For games with a negative edge, note the phrase session budget rather than bankroll. You cannot grow a bankroll playing roulette; the expected value is negative on every spin. What you can do is decide how much entertainment you want to buy, set that amount as the session budget, and size bets so the session is likely to last as long as you want it to. At a 2.70% edge and €10 even-money bets, 50 spins an hour costs an expected €13.50 per hour — the standard deviation over that hour is about €71, so individual hours will be far more volatile than the average suggests.

What a bankroll actually is

Your bankroll is a fixed sum of money, separated from everything else you own, whose only job is to fund your play. Three properties define it:

• It is money you can lose in full without missing rent, a bill or a savings goal. If losing it would change a real-life decision, it is not a bankroll — it is household money at risk.
• It is physically separate: a dedicated e-wallet, a separate bank account or a crypto wallet used for nothing else. Mixing it with everyday money is how "just one more deposit" happens.
• It has rules attached for when you move up in stakes, when you move down, and when you stop for the day. The rules are written before you play, not decided mid-session.

The purpose of all the numbers below is to make one thing true: the biggest losing run you are statistically likely to face should never consume more than a fraction of the bankroll. Once that holds, short-term results stop mattering emotionally, and you can make decisions on expected value rather than fear.

Buy-in rules for poker cash games

The standard advice — 20 buy-ins for cash games — is a starting point, not a law. The correct number depends on your win rate, your standard deviation and how much risk of going broke you are willing to accept. Happily, there is a formula for that.

The risk-of-ruin formula

For a player with a win rate WR (in big blinds per 100 hands) and a standard deviation σ (also per 100 hands), the bankroll B needed to keep the risk of ruin at a chosen level r is:

B = (σ² ÷ (2 × WR)) × ln(1 ÷ r)

A typical 6-max no-limit Hold'em player has σ of around 80–100 bb/100. A decent low-stakes winner makes about 5 bb/100 after rake. Plugging those in:

Two lessons fall out of the table. First, the classic "20 buy-ins" is only safe if you are a genuinely solid winner — and most players overestimate their win rate. Second, a marginal winner needs roughly three times the bankroll of a strong one. If you do not know your win rate over at least 50,000 hands, assume you are marginal and use 40 buy-ins. That is why we recommend 20–40 buy-ins for cash games and lean towards 40 until your data says otherwise.

Moving up and moving down

Use a shot-taking rule: when your bankroll reaches the buy-in requirement for the next level, you may play there, but if you lose a pre-agreed number of buy-ins (typically 3–5) at the new level, you move back down without argument. Moving down is not failure; it is the rule doing its job.

Buy-in rules for tournaments

Tournament variance is brutal because the payout structure is top-heavy: in a typical 1,000-runner event, around 85% of entries win nothing, and most of the prize pool sits in the final three places. A good player with a 30% return on investment (ROI) can go 100 tournaments without a final table and still be playing well.

One detail that catches players out: tournament decisions near the money are governed by the Independent Chip Model, not raw chip counts. Before you call an all-in on the bubble, run the spot through the poker ICM calculator — a call that is +EV in chips is often −EV in money, and bankroll leaks often hide there.

The Kelly criterion, explained without the algebra

The Kelly criterion answers a simple question: given an edge, what fraction of my bankroll should I bet to grow it as fast as possible without going broke? For a bet that pays b to 1, with a win probability p and a loss probability q = 1 − p, the Kelly fraction is:

f* = (b × p − q) ÷ b

Worked example

Suppose you have found an even-money bet (b = 1) that wins 55% of the time. Then f* = (1 × 0.55 − 0.45) ÷ 1 = 0.10. Kelly says bet 10% of your bankroll. Now suppose the bet pays 2 to 1 (b = 2) and wins 40% of the time: f* = (2 × 0.40 − 0.60) ÷ 2 = 0.20 ÷ 2 = 0.10 again — a smaller win rate, but better odds, so the same stake. If a bet wins 50% at even money, f* = 0: there is no edge, so bet nothing. If p is below 50% at even money, f* is negative, which means the correct stake is zero and the correct action is to be the bookmaker.

Why nobody bets full Kelly

• Full Kelly assumes you know your edge exactly. In practice you estimate it, and overestimating the edge by even a little makes full Kelly over-bet, which is worse than under-betting by the same amount.
• Full Kelly produces enormous swings: there is a 50% chance of losing half your bankroll at some point before doubling it.
• The usual compromise is half Kelly or quarter Kelly: you give up a little growth for a large reduction in volatility. Half Kelly delivers about 75% of the growth rate with half the variance.
• For casino games with a negative edge, Kelly is negative. The theory is telling you the same thing the house edge does: there is no stake size that turns roulette or baccarat into an investment.

For poker, Kelly is applied indirectly: the risk-of-ruin formula above is derived from the same growth maths. A cash player using 40 buy-ins at 5 bb/100 is, roughly, playing at about quarter Kelly — conservative, which is exactly where you want to be when your win-rate estimate is uncertain.

Variance: what a normal losing run looks like in numbers

Most players have never seen what their own variance looks like written down, which is why a normal downswing feels like a crisis. Here is what the numbers say for three common situations.

Poker cash game: 10,000 hands

A player winning 5 bb/100 with σ = 80 bb/100 plays 10,000 hands (100 blocks of 100). Expected profit: 100 × 5 = +500 bb. Standard deviation over 10,000 hands: 80 × √100 = 800 bb. So the 95% range of results is 500 ± 1,600 bb, i.e. anywhere from −1,100 bb to +2,100 bb — eleven buy-ins lost to twenty-one won. A winning player loses over a 10,000-hand sample roughly one time in four. If you have not played 10,000 hands since your last deposit, you have learned almost nothing about your skill from your results.

Blackjack: a four-hour session

You play basic strategy at €10 a hand, 100 hands an hour, house edge 0.5%. Over 400 hands the expected loss is 400 × €10 × 0.005 = €20. But the standard deviation of a blackjack hand is about 1.15 bets (doubles and splits inflate it), so over 400 hands σ ≈ 1.15 × €10 × √400 = €230. A session finishing anywhere between about −€480 and +€440 is within two standard deviations. That is why a €100 session budget at €10 bets runs out so often: the budget is less than half of one standard deviation.

Tournaments: a 100-buy-in downswing

A 20% ROI player in 1,000-runner events cashes about 15% of the time. Simulations of this profile regularly show downswings of 100–150 buy-ins and break-even stretches of 500–1,000 tournaments. This is not bad luck; it is the expected texture of a top-heavy payout structure. The 200–300 buy-in recommendation exists precisely so that a 100 buy-in downswing costs you a third of your roll, not all of it.

Stop-loss, stop-win and session rules

Mathematically, a stop-loss does not change your expected value: the cards do not know you are losing. Its job is psychological and practical — it removes the decision from a moment when you are least able to make it. Use these rules and write them down.

1. 1

   Set a session stop-loss in units before you sit down

   Cash poker: 3 buy-ins. Table games: the full session budget, and never reload. Sports: no more than the day's planned stake. When you hit it, you close the client. Not "one more orbit".

2. 2

   Set a time limit as well

   Fatigue degrades decision quality faster than most players notice. Two to four hours for poker, one to two hours for fast table games. Take a ten-minute break each hour.

3. 3

   Use a soft stop-win

   Quitting while ahead does not lock in long-term profit, but if you notice you are playing looser because "it's the house's money", that is a leak. Bank a portion of the win and tighten up, or stop.

4. 4

   Never chase within a session

   Increasing stakes to recover a loss is the single most destructive behaviour in gambling. If you feel the urge, that is the signal the session is over.

5. 5

   Review weekly, not hourly

   Check your bankroll, win rate and unit size once a week. Adjust stakes according to your move-up/move-down rule, not according to how yesterday felt.

Tilt: the bankroll leak that is not in the maths

Tilt is any emotional state — anger, frustration, boredom, overconfidence — that makes you deviate from the strategy you would use when calm. It turns a 5 bb/100 winner into a 10 bb/100 loser for as long as it lasts, and it is the main reason real bankrolls fail faster than the risk-of-ruin formula predicts.

• Recognise your triggers. Common ones: a bad beat for a large pot, a run of coolers, being outdrawn by a player you consider worse, or simply being tired.
• Have a pre-committed response. "If I lose two buy-ins in an hour, I stand up for fifteen minutes" works because it is decided in advance.
• Shrink the unit, not the discipline. Some players move down a level when they feel tilt approaching. That is a legitimate tool; moving up is not.
• Separate results from decisions. Review hands by asking "was the decision correct given what I knew?", not "did I win?". Results-oriented thinking is the fuel of tilt.

Keeping bankroll and life money apart

The most important rule in this guide is the least mathematical. Your bankroll must be money whose loss would not matter. Practically, that means:

• Fund the bankroll once, from disposable income, and do not top it up from salary when it shrinks. If it runs out, that is information, and the answer is to study or to stop, not to redeposit.
• Keep it in a separate account or wallet. Never keep a casino account linked to a card you use for groceries.
• Withdraw profits on a schedule. A bankroll that only ever grows on the site is not real money yet; a bankroll that pays for a holiday is.
• Set deposit limits on every platform you use. Good operators let you set daily, weekly and monthly limits; our reviews note which ones make these tools easy to find.
• If you would be embarrassed to tell a partner or friend how much is in the bankroll, it is too large relative to your life.

Putting it together: a worked bankroll plan

Suppose you set aside €1,000 that you can genuinely afford to lose and you want to play online no-limit Hold'em cash games. Using 40 buy-ins, your buy-in is €25, which is 100 big blinds at €0.10/€0.25 — in other words NL25. Your session stop-loss is 3 buy-ins (€75). You move up to NL50 when the roll reaches €2,000 (40 × €50), and you move back to NL25 if you lose 5 buy-ins at NL50. You track hands, and after 50,000 hands you revisit the risk-of-ruin table with your measured win rate. If the same €1,000 were for tournaments, 200 buy-ins gives a €5 buy-in; for blackjack, it is not a bankroll at all but a pool of session budgets — say €100 a session at €1 bets, which is 100 units and comfortably survives a normal four-hour variance swing.

From here, the natural next step is the game-specific strategy: poker strategies for how to build the win rate your bankroll is waiting for, blackjack strategies for keeping the edge at half a percent, and the wiki for definitions of variance, ROI, ICM and standard deviation.