Nobody can promise you will not go broke playing poker. Skill changes the shape of the distribution. It does not delete the left tail. Thirty buy-ins for cash, a hundred for multi-table tournaments, fifty for sit-and-gos — those are 2026 starting points, not a sacrament, not a shield, not a personality. They are harm reduction. Treat them like a seatbelt. Seatbelts do not make you a good driver. They make the crash survivable more often.
If you want a calculator that turns a number into a suggested buy-in, use the bankroll tool. Then come back here for the part a slider cannot do: the written move-down, the shot, the downswing that lasts longer than your ego, and the moment you should stop treating "one more bullet" as a study plan.
The numbers, as guidance
| Format | Starting point | Written move-down | Why it is fatter than your uncle's advice |
|---|---|---|---|
| Cash, 100bb | ~30 buy-ins | ~20 buy-ins | Win rates are thinner than 2010. Rake is not. A 4 bb/100 edge with an 80 bb/100 standard deviation will still punch you in the mouth for 15 buy-ins. |
| MTTs | ~100 buy-ins | ~60 buy-ins | You can play well for a month and cash nothing that matters. The right tail pays you. The left tail is silence. Silence needs a roll. |
| SNGs / Spins | ~50 buy-ins | ~30 buy-ins | Lower variance than a 2,000-player field, higher than a cash table. Fifty is a belt. Twenty is a dare. |
Move-down means you already wrote the number before you lost. "I'll know when it feels right" is how people grind 12 buy-ins at a stake they have no business in, because last month they had 40 and a story. Write it. Put it on a note. When the roll tags the number, you move down that night, not after one more session to "get even." Getting even is not a strategy. It is how stuck sessions become stuck weeks.
Move-up is the other note. Cash: you do not move up from 30 to a bigger stake at 31. You move up when you can play the new stake at ~30 and still have a move-down that does not dump you into a game you cannot beat. If NL50 is the stake you crush and NL100 is a shot, the shot gets a budget. The crush game stays available. That is game selection wearing a bankroll hat.
A 20-buy-in hole is not exotic
NL100, 100bb buy-in = 100 big blinds. You win 4 bb/100 hands (respectable, not a fable). Standard deviation 80 bb/100 (also not exotic for 6-max).
Over 50,000 hands: expected profit 4 × 500 = 2,000 bb, about 20 buy-ins. Standard deviation of that sample: 80 × √500 ≈ 80 × 22.4 ≈ 1,790 bb, about 18 buy-ins.
A result 1.5 standard deviations below expectation is 2,000 − 2,685 ≈ −685 bb, call it −7 buy-ins, after a volume that took a serious winner months. A 2-SD underperformance is 2,000 − 3,580 = −1,580 bb, about −16 buy-ins. That is still a winner's distribution. You do not get to opt out because you "played well." You did play well. The sample said no for a while.
If you sat down at this stake with 20 buy-ins and no move-down, that 2-SD month put you in the wreckage. If you sat down with 30 and a written move-down at 20, you are at NL50 grinding your actual edge instead of taking a pride-based shot with 11 bullets. That is the entire argument. It is not mystical. It is arithmetic plus a promise you make to yourself when you are not stuck.
Tournaments are fatter tails. A 15 bb/100 ROI in a mid-stakes field is a good player. You can still go 80 events without a score that matters. Eighty events at 100-buy-in discipline is 0.8 of a roll in entry fees, not counting the times you did cash. Eighty events at 25-buy-in "discipline" is how a winner becomes a former winner over a quiet quarter. The field did not get tougher in one week. Your roll got thinner and your decisions got needier. Needy is not a style. Needy is a leak with a calendar.
Risk of ruin, without the textbook fog
Risk of ruin is the probability that you hit zero — or hit your "I have to stop" number — before the edge has time to show up. Edge down, variance up, bankroll down: ruin risk up. That is the whole function. You do not need a Greek letter to use it.
Cash: your edge is small in bb/100, your variance is large in bb/100, your sample per session is small. Ruin is a real visitor at 15 buy-ins even if you are actually a winner. MTT: your edge might be larger in ROI terms and the variance is a cliff. Ruin at 40 buy-ins is not a freak accident. It is what the distribution does when the right tail takes a quarter off.
Two ugly truths. First, you do not know your win rate as well as you think, so any ruin number you compute with "I'm 8 bb/100" is fan fiction if your sample is 40,000 hands of mixed games. Second, going broke is not only hitting zero. It is hitting the number where you start taking shots you did not plan, chasing, hiding sessions, or borrowing. If the roll is gone in spirit, the last chips are a formality. Harm reduction means you never give those last chips a vote.
This is not a promise you cannot go broke. Winners go broke when they play too high, play too long stuck, or play a game they do not beat. The guidelines lower the odds. They do not sign a contract with the universe.
Downswing psychology is part of the math
The distribution above is why a downswing can last longer than your narrative about who you are. You will want to change something. Sometimes you should: the table is tougher, you are playing tired, you added mix-max zoom at midnight, your continuation bets are 2016. Sometimes you should not: you lost with Ace-King versus Queens three times in a week and now you "don't 3-bet as much." That is results-oriented vandalism. That's a cooler, not a leak, and you still spewed if the next adjustment is to start calling river bets with third pair because you are "due."
Write the strategy changes you are allowed to make while stuck: study a leaked node, sleep, move down, stop at a loss cap. Write the ones you are not: inventing a new river philosophy, jumping a stake to "win it back faster," adding games you do not have a graph for. Stuck is not a read. Stuck is a status. GTO versus exploit still applies. Your "exploit" cannot be "I am down."
A session cap is allowed. A stop-loss of two or three buy-ins in cash, or a hard number of bullets in tournaments, is not superstition if the point is to keep a tired brain off the table. It is superstition if you think the cards owe you a window after you leave. They do not. You leave because your decisions degrade, not because the deck has a memory.
Shot-taking, as a budget, not a dare
A shot is a pre-planned, pre-sized attempt at a higher stake or a tougher field, with a return ticket. Example: you have 45 NL50 buy-ins. You peel 8 of them as NL100 bullets. If they are gone, you are still on 37 NL50 and you go back without a speech. If they grow, you rewrite the plan. What is not a shot: waking up at NL200 because last night's heater felt like a new identity.
Conditions for a shot that is not a punt:
- The lower game is still there, and you still beat it.
- The shot has a bullet count written down. Not "we'll see."
- The game you are shooting is actually selected — a good table, a soft field — not just a bigger sign on the same grind.
- You can fail the shot without skipping rent, hiding a transfer, or lying to whoever shares your books.
If those fail, it is not a shot. It is an escape. Escapes belong on the responsible gambling page, not in a strategy note. There is a difference between variance and using the table to change how you feel. If you cannot tell which one you are doing, you are doing the second.
Game selection is bankroll management
Thirty buy-ins of a terrible game is a slower bust than five buy-ins of a terrible game, not a strategy. The roll exists to let you say no. No to the 7-max zoom table full of nits and regs who have you tagged. No to the live 2/5 that is actually 2/5/10 with a mandatory straddle and three pros in seats 3, 5, and 8. No to the 1,000-player turbo at a buy-in you only play because the overlay tweet was funny.
Yes to the live 1/3 with seven recreational stacks and a lineup you can describe in a sentence. Yes to the cash table where two seats are leaving money in the middle. Yes to the tournament you beat at a buy-in your roll can repeat. Volume is not a virtue when the volume is a tax. Live versus online changes the texture of that tax. It does not change the rule: pick games your roll can survive and your edge can find.
Street by street: the cooler, then the extra bullet
This is not a hand about range construction. It is a hand about what people do with the next orbit. Live 1/2, 200-point stack, you have 2,400 behind. You have been winning for six weeks and you brought 40 buy-ins to this room this year. Tonight you sat with 300 because the lineup looked good. That part was correct.
Preflop — $1/$2, 150bb, UTG opens
UTG, a competent rec, opens 12. You 3-bet to 40 on the button with Kings. Big blind, unknown, cold-calls. UTG calls. Pot 123. This is a dream. It is also how people go broke when the dream has Aces on the other side and they do not have a next-bullet plan.
Flop
BB checks. UTG checks. You bet 70. BB raises to 200. UTG folds. You call. You should. This is a King-high, two-club board and you have the overpair without a club. Sets exist. Flush draws exist. Queen-x exists. You are not folding flop. You are also not shoving over the raise unless you have a specific read that this human never has a set. You do not.
Turn
Pair the board. BB shoves 310 into 493. This is a miserable, standard spot. Against a range of made Queens, sets, and some flush draws, you are often flipping or slightly behind. You call. They show Ace-Ace. That's a cooler, not a leak. You stacked off with Kings in a 3-bet pot versus a cold-caller who had Aces. The poker was fine.
The next orbit — this is the article
You rebuy 300 you did not plan to fire. You 3-bet Queen-Jack suited in a spot you would have flatted an hour ago. You empty the clip on a ten-high board because you are "not going to get run over." They show Ten-nine. You are out a second bullet. The Kings were not the leak. The extra bullet plus the invented 3-bet was the leak. If that second 300 was inside a written shot, fine — still a bad hand. If that second 300 was rent-adjacent, or the 19th buy-in of a 20-buy-in floor you had promised to honor, you did not "gamble up." You broke the only rule that was protecting you from the distribution we already did the math on.
Bankroll management is the decision before the Kings, and the decision after them. The cards in between are poker. Do not confuse those three files.
Leak: the heater promotion, the silent roll
You ran 12 buy-ins up in three weeks and bought in 40% higher as a new identity. Then the mean-reversion month arrived, which is just the other half of the same standard deviation, and you had no move-down because that would "feel like a demotion." Demotion is a word from jobs. This is inventory. You sell inventory when the price is wrong. The other leak is not tracking: mixing book money, bonus money, and the card in your wallet until you cannot say how many buy-ins you have. If you cannot say the number, you do not have a bankroll. You have a fog. Fog is where ruin lives.
Deviate: take the shot when the table is actually good
The guideline is 30. The table in front of you is the best 2/5 you will see this quarter, you have 28 NL200 buy-ins in the cash roll, and you can fire two bullets without tagging your move-down. Take it. Write "two." Leave when they are gone or when the lineup changes. The opposite deviation is also allowed: you have 50 buy-ins for a stake you do not beat, because the regs there are better and the rake is hungrier. Move down without waiting for the guideline to save you. Edge first, then inventory. A fat roll in a negative-EV game is a countdown.
What the roll is not
It is not a scoreboard of your worth. It is not a reason to play when you should sleep. It is not a substitute for game selection. It is not a talisman against a 2-SD month. It is inventory for a gambling game you are trying to play with an edge. If the edge is gone, stop buying inventory. If the game is the only thing that makes a day feel like a day, that is no longer a bankroll conversation. Read responsible gambling before you fire another bullet. Friends, family, and the NCPG helpline are more useful than a strategy article when the fire is the point.
Keep that paragraph away from the cute idea that a magazine can fix a compulsion with a table of buy-in counts. It cannot. The table of counts is for people who are still choosing.
A 30-minute bankroll session
- Count the poker money only. Not the rent. Not the "I could take it from savings." The poker money.
- Divide by the buy-in you actually play. Cash in 100bb units. MTTs in entry units including the juice you really pay.
- Write the move-down number. Put it where you will see it when you are stuck, not when you are writing a plan in a good mood.
- If you want a shot, write the bullet count and the return stake.
- Open the bankroll tool and see if your ego and the slider agree. If they do not, the slider is not the one tilting.
That is the work. It is dull. Dull is what keeps a winner in the game long enough for the 4 bb/100 to show up. Flashy is the extra bullet after Kings.
FAQ
Is 30 buy-ins a law?
No. It is a cautious 2026 cash starting point for a 100bb game with real rake and real opponents. If you beat a live 1/3 for a fat hourly with a lineups-first habit, you can live a bit leaner. If you play 6-max zoom against regs, live fatter. If you were about to treat 30 as a dare to go lower, you missed the point. The number exists to keep you off the cliff, not to give you a running start at it.
Why do tournaments need ~100 buy-ins?
Because a good player can go a long time without a score that moves the roll, and because one score should not force you into a bigger buy-in just to feel the same rush. Sixty as a move-down is the "you are now in the weather" line. Under 40 in a serious MTT grind is how winners disappear between series.
Can I mix cash and MTT rolls?
You can, if you still count in the most demanding unit. A combined pile that is 80 MTT buy-ins and also 25 cash buy-ins is not both. It is a cash roll that is already near a move-down and an MTT roll that is thin. Separate them or treat the whole thing as the stricter format. Mixing without counting is the fog again.
What if I only play live once a month?
Then variance per hour is the same and variance per month is lumpier. You still want a pile that can absorb a three-buy-in night without touching rent. You still want table selection. You do not get a pass on the math because the felt is romantic. If the monthly trip is entertainment, budget it as entertainment and stop calling it a roll. Entertainment money is allowed. Pretending entertainment money is a 30-buy-in grinding stake is how people come home smaller than they meant to.
Does a downswing mean I should quit?
A downswing means you should compare your actual decisions to a baseline, check whether you still have an edge in the games you are in, honor the move-down, and sleep. It does not automatically mean you are a loser. It also does not automatically mean you are a winner who is "due." Due is not a number. If poker is making the rest of your life smaller, quitting is on the table whether you are up or down. That question lives on the responsible gambling page, not in a win-rate spreadsheet.
Where does rakeback or a reload fit?
It doesn't, here. This is not a store. Lower rake is a better game; that is table selection. We are not selling you a deal, a code, or a cashier. If a room's rake is unplayable at your stake, move tables or move down. Do not wait for a promotion to make a minus-EV game plus.