Bankroll Management: The #1 Rule Most Players Ignore

Most recreational players who go broke do not go broke because they play bad poker. They go broke because they play a stake their bankroll cannot survive, run into a completely ordinary downswing, and have no rules for what happens next. This article gives you the rules — and, unlike most bankroll articles, the simulated numbers that justify them.

How many buy-ins do you need?

The golden rule: never play a stake where you hold fewer than 30 buy-ins. A buy-in in cash games is 100 big blinds — the standard maximum sit-down at most online tables. Thirty is not a superstition; it is roughly the depth that survives the downswings a genuine winner should expect (the data is two sections down).

Thirty is the floor; fifty is comfortable. Push toward fifty if you play 6-max with a modest win rate, tilt easily, or cannot redeposit. The cost of being over-rolled is slower stake progression; the cost of being under-rolled is the whole bankroll.

What does a normal downswing actually look like?

We ran 20,000 simulated 50,000-hand and 100,000-hand samples through the same model our variance calculator uses (each hand drawn from a normal distribution matching the stated win rate and standard deviation). Here is what happens to players who are genuinely winning:

Read that middle row again: a true 3bb/100 winner — a solid regular — loses money over a full 50,000-hand sample about one time in four, and more often than not experiences a 20+ buy-in peak-to-trough downswing somewhere inside it. Over 100,000 hands the median worst downswing grows to roughly 27 buy-ins, because more hands means more chances for the bad stretch to appear. This is what 30-50 buy-ins is protecting you from: not bad play — normal luck.

We have since scaled this analysis to 70 billion simulated hands: 81% of true 3bb/100 winners hit a 20+ buy-in downswing somewhere in a 100,000-hand career, and 39% hit 30+ buy-ins.

What standard deviation should you plug in?

Both simulation rows above depend on the σ you feed the model, and σ is mostly a property of the game format, not of you. These are the format defaults our win rate calculator uses:

The practical read: the 30/50 buy-in rule in this article is calibrated for 6-max and full-ring hold'em, where σ sits in the 80-110 band. The moment you switch to a higher-variance format, the same win rate produces much deeper swings — heads-up and PLO players routinely keep 50-100 buy-ins for exactly this reason. If you play a mix, budget for the swingiest format in the mix, not the average.

What per-session rules keep a roll alive?

The stop-loss is the important one. Statistically, the cards do not know you are stuck — but you do, and decision quality degrades fast when losing. The stop-loss is a circuit breaker between an ordinary 3 buy-in bad night (which the variance table above says is routine) and a tilt-driven minus-15 that actually dents a bankroll. Tilt destroys more bankrolls than variance ever has.

When should you move up — and how much proof do you have?

Three conditions must all be true before you climb a stake:

• ✓ You have 30+ buy-ins for the new stake — going from NL10 to NL25 means $750, not $300.
• ✓ You are winning over at least 20,000 hands at your current stake.
• ✓ You are moving up on a plan, not on a heater and not chasing a loss.

A dose of honesty about the second condition: 20,000 hands is an evidence bar, not proof. In the simulation above, even a true 3bb/100 winner finishes a 20,000-hand stretch in the red about a third of the time — and by the same token, a break-even player books winning 20k stretches constantly. To pin down a 3bb/100 win rate at 95% confidence (σ = 100) takes on the order of 400,000 hands. Our win rate calculator runs that confidence-interval math on your actual sample. This is exactly why the bankroll cushion, not the win-rate estimate, does the real protecting.

How do you take a shot without risking the roll?

• → Risk no more than 5% of your bankroll on a shot at the next stake.
• → Pre-commit a 2 buy-in stop-loss at the new stake. Lose them, drop straight back. No exceptions, no "one more orbit."
• → Shoot when tables are soft, not when you are bored or stuck.

If the shot sticks — you are winning at the new stake over a few thousand hands — start building the full 30 buy-in roll for it. If it fails, it cost a planned, survivable 5%. Shot discipline converts moving up from a gamble into a process.

Why is moving down a strength?

When your roll drops below 20 buy-ins for your current stake, move down immediately. Every downswing in the simulation table kept going before it turned around; at 20 buy-ins you no longer have the depth to let it. Dropping down lets you rebuild at a stake your roll dominates — with calmer decisions, against softer opponents. The players who never go broke are not the ones who never run bad; they are the ones who move down on time.

Do tournaments need a different bankroll?

Yes — a much bigger one, and the same simulator shows why. Run the variance calculator's MTT-grinder preset (8bb/100 with a σ of 200bb/100 — double the win rate of our cash examples, but double the volatility too) over 50,000 hands: about 18% of samples still lose money, the median worst downswing is around 38 buy-ins, and the unlucky 1-in-20 tail runs roughly 75 buy-ins deep. Downswings scale with σ, and tournament σ is brutal because most of the prize pool hides in top-three finishes you rarely reach.

That is where the standard advice of 100+ buy-ins for tournaments comes from — it is not conservatism, it is the cash-game logic applied to a σ twice as large. Count your roll in average buy-ins (a $5 average means $500+), and if your schedule mixes large-field events, hold more still: field size raises variance faster than it raises your edge.

How hard does rake hit a micro-stakes bankroll?

Harder than almost anyone realizes. Our rake comparison tool models the cost from published rake structures (5% rake at most rooms, ~16 raked pots per 100 hands, typical micro pot sizes): the gross cost lands between roughly 7.5 and 9.4 big blinds per 100 hands at NL2, narrowing to about 6.8-7.9bb/100 by NL25 as caps start to matter. Compare that to win rates: a strong NL2 player beats the game for maybe 8bb/100 afterrake — the room's cut is on the same order as the winner's entire profit.

Two practical consequences. First, rakeback is not a bonus — at the micros it is a material share of any realistic win rate, and room choice can swing your net result by 2-3bb/100. Second, rake relief is one of the honest arguments for moving up when your roll and skill allow: the same 5% bite shrinks relative to pot sizes as caps engage at higher stakes.

A 30-day bankroll reset

• D 1-3 Calculate your true roll. Open a separate poker e-wallet. Move the bankroll there — only the bankroll — and set your stake rules in the bankroll manager.
• D 4-7 Drop one stake below your usual. The goal is to win small at a level your roll dominates.
• D 8-30 Log every session in the session journal: stake, hands, result, mood, biggest mistake. Review weekly. Enforce the 3 buy-in stop-loss without exception.

Most players who run this reset discover their problem was never card-sense — it was session management. The stop-loss alone eliminates the worst of it.

Common questions