Guide 5 of 8 / 8 min read
Bankroll Management: Unit Sizing, Staking and Downswings
One bankroll, sized in units
- Bankroll
- 2,000
- Unit size
- 1%
- One bet
- 20
A 2,000 bankroll at 1% units means a standard bet is 20. The bankroll holds 100 of them: at or above the 100 units many bettors want as a rough floor.
A bankroll is money you have set aside for betting and can afford to lose entirely without it changing your life. Keeping it separate from your day-to-day finances is the whole point: it lets you make correct long-run decisions without a single bad month forcing your hand.
Written and reviewed by Alexander Korsager · last reviewed September 29, 2026 · corrections
The short version
Count in units, not currency
A unit is a fixed fraction of the bankroll, commonly 1%, so a 2,000 bankroll makes a standard bet 20.
Losing runs are expected
A bettor with a 2% edge at even money still finishes below zero after 500 bets about 31% of the time.
Buy-in rules assume moving down
Held at a stake that never drops, 20 buy-ins carries a 22.7% risk of ruin for the guide's example winner.
Bankroll management is the set of rules that decide how much of that money goes on each bet. Done well, it keeps you in the game through the inevitable losing runs so that your edge has time to show up. Done badly, it busts a winning bettor before variance evens out.
01 / 05
Setting a unit
Express stakes in units rather than currency. A unit is a fixed fraction of your starting bankroll, commonly 1%, sometimes up to 2% for bettors confident in their edge, less for beginners. A 2,000 bankroll at 1% units means a standard bet is 20.
2,000 × 1%20
One unit out of the 100 the bankroll holds.
- At 1%
- 20
- At 2%
- 40
Thinking in units keeps your staking proportional and stops you chasing. A losing day is "down four units", not "down 80", which is easier to keep in perspective.
- In units
- 4 units
- In currency
- 80
4 × 2080
02 / 05
Flat versus percentage staking
Flat staking bets the same currency amount every time, recalculated occasionally. It is simple and its results are easy to read, but it does not adjust as your bankroll moves.
- Bankroll left after 10 losses in a row
- 1,800
Percentage staking bets a fixed fraction of your current bankroll, so stakes shrink automatically during a downswing and grow as you recover. It compounds gains and cuts losses faster, at the cost of never quite recovering to the same stake without winning first. Most disciplined bettors use flat staking within a session or week and resize units periodically.
- Bankroll left after 10 losses in a row
- 1,808.76
03 / 05
Downswings are normal
Even a genuine 55% bettor at even money will have losing runs of eight or ten in a row over a long enough sample. A separate case: a bettor carrying a smaller 2% edge at even money still finishes below zero after 500 bets about 31% of the time, roughly one run in three. Neither result is a sign the edge is gone, it is variance doing exactly what the maths says it will.
500 bets at a 2% edge, even money
The remaining 69 runs in 100 finish at or above zero over the same 500 bets. Neither result says the edge is gone, it is variance doing exactly what the maths says it will. Methodology and sourcing below.
The practical defenses are a big enough bankroll relative to your unit, stakes that do not balloon after losses, and a written plan you follow when you are tilted rather than one you invent mid-slide.
04 / 05
How much bankroll is enough
It depends on your edge and your variance, not just your edge. A low-variance approach needs fewer units of bankroll for the same risk of ruin than one chasing long-shot multipliers. As a rough floor, many bettors want at least 100 units before betting 1% stakes, and treat anything under 50 units as under-rolled for the stakes.
- One unit of 100
- 1%
- One unit of 50
- 2%
Both assume a given edge and variance, so neither is a safe count for every strategy.
05 / 05
Poker counts the same thing in buy-ins
A poker bankroll is quoted in buy-ins rather than in units, which is the same idea at a different scale. A buy-in at a no-limit cash game is usually 100 big blinds, so the common advice to hold 20 to 30 buy-ins is advice to hold 2,000 to 3,000 big blinds. Tournament conventions run far higher, commonly cited at 100 buy-ins or more, because tournament results are concentrated in a small share of finishes and the swings are correspondingly longer.
No-limit cash games
- Tournaments, in buy-ins
- 100+
Those counts quietly assume something the arithmetic does not. Put the same player through a fixed-stake risk of ruin, the formula published in Chen and Ankenman's The Mathematics of Poker, and a winner at 3 big blinds per 100 hands with a standard deviation of 90 carries a 22.7% risk of ruin on 20 buy-ins and 10.8% on 30. About 40 buy-ins is what a 5% risk actually asks for at those numbers. The rule of thumb is not a safer figure hiding a stricter one, it is a figure that works because the player following it drops a level rather than playing the last of the roll at the old stake.
- 22.7%20 buy-ins
- 10.8%30 buy-ins
- 5%About 40 buy-ins
A winner at 3 big blinds per 100 hands, standard deviation 90.
That makes moving down part of the bankroll rule rather than a failure of it, and it is doing most of the protecting. It also means a poker bankroll cannot be sized once and left. Moving up a level halves what the same money is worth counted in big blinds, and it resets both inputs, since the standard deviation and the win rate belong to the game being played rather than to the player.
The win rate and the standard deviation belong to the new game too.
Check yourself
Can you size a bankroll now?
1 / 3
A 2,000 bankroll at 1% units makes a standard bet of what?
2 / 3
After 10 losses in a row, which plan asks for a smaller next stake?
3 / 3
Held at a stake that never drops, what risk of ruin does 20 buy-ins carry for a winner at 3 big blinds per 100 hands with a standard deviation of 90?
Pick an answer to see the working. Nothing is scored or saved.
Try it
Put your own numbers through it
- Bankroll Calculator Size your stake per bet
- Kelly Stake Find your optimal bet size
- Risk of Ruin Calculator Find the odds of going broke
- Variance Simulator Watch a downswing play out
- Problem Gambling Screening (PGSI) The nine-question PGSI, scored as published, with support resources beside your result. Nothing you enter is saved.
FAQ
- What percentage of my bankroll should I bet?
- A flat 1–2% per bet is the common range. Going above roughly 5% meaningfully raises your risk of ruin during a normal losing streak, even with a real edge.
- How many buy-ins should a poker bankroll be?
- The convention for no-limit cash games is 20 to 30 buy-ins, and for multi-table tournaments it is far higher, commonly cited at 100 or more. Both assume a player who moves down a level when the roll drops. Held to a stake that never changes, 20 buy-ins is a 22.7% risk of ruin for a winner at 3 big blinds per 100 hands with a standard deviation of 90, and 40 buy-ins is roughly what a 5% risk asks for. The risk of ruin calculator solves it from your own win rate and standard deviation, which is the version that matters here, because the standard deviation enters the formula squared.
- Should I increase stakes after a loss to recover?
- No. Loss-chasing progressions like the Martingale move risk into a fat tail without changing your expected value. Keep stakes proportional to your current bankroll and let the edge work.
- How often should I resize my unit?
- Periodically rather than after every bet, monthly, or after the bankroll moves by a set amount such as 25%. Constant resizing turns flat staking into percentage staking with extra admin.
Where these figures come from
The poker risk-of-ruin figures in 'Poker counts the same thing in buy-ins' and the FAQ restating them are this site's own Risk of Ruin tool: continuousRoR(winRate100, sd100, bankrollUnits) = e^(-2 · winRate100 · bankrollUnits / sd100²) in src/tools/risk-of-ruin/math.ts, evaluated at the guide's own stated inputs (3bb/100 hands win rate, 90bb/100 hands standard deviation, buy-ins converted to big blinds at the standard 100bb-per-buy-in convention this section states). 20 buy-ins (2,000bb) returns 22.73%, 30 buy-ins (3,000bb) returns 10.84%, and bankrollForRoR at a 5% target returns 40.4 buy-ins, matching the guide's '20 buy-ins... 22.7%... 30... 10.8%... About 40 buy-ins' exactly. These are the same functions the risk-of-ruin page itself uses for its edge-model figures, and math.test.ts checks bankrollForRoR is the true inverse of continuousRoR. The underlying formula is the fixed-stake risk-of-ruin result published in Chen and Ankenman's The Mathematics of Poker (2006), which the guide's own body already names rather than presenting the maths as this site's own invention. The '31% of the time, roughly one run in three' figure in 'Downswings are normal' is this site's own EV Calculator: probabilityForEdge(0.02, 2.00) recovers the win probability that reproduces a 2% edge at even money, and placedBetsOutlook's probabilityBehind is an exact forward walk over the binomial win count across 500 bets at that probability, not a simulation, from src/tools/ev-calculator/math.ts. Even money is a stated price choice, matching the sentence immediately before it ('at even money'), because the claim is price-dependent and the guide's sentence names no price of its own; src/guides/figures/bankroll-downswing-data.ts records the choice and the reason. risk[2]'s 'genuine 55% win rate over 500 bets' names a win rate, not the 2% edge this figure is about (a 55% win rate at even money is a 10% edge, evPerUnit(0.55, 2.00) = 0.10), so the two sentences describe different, correctly labelled scenarios rather than one restating the other. The '20 to 30 buy-ins' cash-game and '100 or more' tournament conventions, and the sports-betting '1-2% per bet, 100 units as a rough floor' figures earlier in the guide, are widely cited industry rules of thumb rather than values this site's own code derives; the guide presents them as conventions, consistently with that.
Verification
Checked against risk-of-ruin's own continuousRoR/bankrollForRoR (src/tools/risk-of-ruin/math.ts) and Chen & Ankenman's The Mathematics of Poker (2006) for the underlying formula; ev-calculator's own probabilityForEdge/placedBetsOutlook (src/tools/ev-calculator/math.ts) for the downswing share: agrees on the 22.7% (20 buy-ins), 10.8% (30 buy-ins) and roughly-40-buy-in (5% target) figures, recomputed directly from the tool's own functions at the guide's stated inputs; the downswing share (about 31% at even money, 500 bets, a 2% edge), recomputed directly from ev-calculator's own functions at the guide's stated price and sample size (verified by src/tools/risk-of-ruin/math.test.ts (continuousRoR, bankrollForRoR); src/guides/figures/bankroll-downswing-data.test.ts (this guide's own downswing figure, re-derived independently from ev-calculator/math.ts), 2026-09-29).
Know the risk
- Treating '20 to 30 buy-ins' as a fixed target you can hold at a stake that never drops. The convention only delivers the risk it advertises if you move down a level when the roll requires it; held to a fixed stake, 20 buy-ins carries a 22.7% risk of ruin for a winner at 3 big blinds per 100 hands with a standard deviation of 90, not the comfortable-sounding figure the rule of thumb implies.
- Treating a rough floor like 100 units as safe for any bettor who clears it. That number assumes a given edge and variance; a higher-variance approach, or a smaller true edge than you believe you have, needs meaningfully more, and there is no single unit count that fits every strategy.
- Reading a losing run of eight or ten in a row as proof an edge has disappeared, and increasing stakes to chase it back. At a genuine 55% win rate over 500 bets, a streak that long is ordinary variance, not a signal to abandon the approach or resize upward mid-slide.
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.