A bankroll is a shock absorber, not a scoreboard. It exists to keep a run of bad cards from ending your ability to play, and that is the whole of its job. It does not know whether you’re good. It does not know whether the game you’re about to sit in is full of regulars who have been studying three-bet pots since you were limping aces.
Which is why the standard answer to “am I ready to move up?” is unsatisfying. Twenty buy-ins, thirty buy-ins, fifty for tournaments and double that if you’re a lunatic: these are answers to a question about survival, not about skill. You can satisfy every bankroll rule ever written and still be a losing player at the next level. You can also be plainly better than the game above you and sit there with twelve buy-ins, which is a different kind of mistake.
Readiness has three separate components and most players only check one. There’s the money, there’s the evidence that you actually beat your current game, and there’s whether your decision-making survives contact with a stake where the numbers feel real. They fail independently, and the third one fails quietly.
The sample you think proves you beat the game almost certainly doesn’t
Here is the arithmetic nobody wants. Your results swing around your true win rate by an amount measured in big blinds per 100 hands, and for no-limit hold’em cash games that spread typically runs near 80 to 100 big blinds per 100 hands. Call it 90. Say you’ve played 25,000 hands at your current stake and you’re up 5 big blinds per 100, which is a strong result.
Now put a confidence band around it. The uncertainty in a measured win rate shrinks with the square root of your sample. Twenty-five thousand hands is 250 blocks of 100. The square root of 250 is about 15.8. So 90 divided by 15.8 gives roughly 5.7 big blinds per 100 as the standard error. The usual 95% band is about twice that in each direction.
Your true win rate, on that sample, sits somewhere between losing 6 and winning 16. You have learned almost nothing. A common rule of thumb says you need roughly the standard deviation divided by your win rate, squared, times 100 hands to get even a one-standard-deviation signal — about 22,500 hands for a 5 bb/100 winner with a 75 bb/100 spread, and one standard deviation is a weak standard of proof. Real confidence takes hundreds of thousands of hands, which most players will never accumulate at a single stake before the games change underneath them.
So stop waiting for your graph to prove you’re a winner. It won’t, not in any reasonable timeframe. Use the graph as a veto rather than a permission slip. If you’re down over 40,000 hands, that’s meaningful enough to pause. If you’re up modestly, that’s consistent with being a small winner, a big winner, or a break-even player who ran fine.
What replaces the sample: evidence from the hands themselves
If results can’t tell you, hand review has to. The useful test isn’t whether you won the pot. It’s whether you can explain your river decision in terms of what your opponent’s calling range actually contains, and whether that explanation survives you looking at it again a week later.
Pull twenty of your biggest losing pots from the last month. Sort them into three piles: coolers where any sane player loses the stack, spots where you were out of ideas and defaulted to a bet, and spots where you knew the right play and did something else. The second pile is a study problem. The third pile is a discipline problem, and it gets worse when the money gets bigger, not better.
Then do the same with twenty pots you won. Players review their disasters and never audit their wins, which is how a bad line that happened to get there goes uncorrected for a year. If more than a couple of your big winners came from an opponent’s gift rather than your plan, your win rate is partly on loan from the player pool, and that loan gets called in at the next level up.
What actually changes one level up, and what doesn’t
Less than people fear, and in a specific direction. The fundamentals don’t change. What changes is the frequency of opponents who punish passive lines.
Concretely: expect more three-bets from the blinds, fewer players calling your opens out of position with weak offsuit hands, and considerably fewer stations paying off three streets with second pair. Your thin value bets get called by better hands more often. Your check-backs get stabbed at more often. The flop continuation bet you’ve been firing at 70% because it kept working starts getting raised.
There’s a compensation, though, and it’s real. Rake is usually capped per pot, so as the blinds get bigger the house takes a smaller slice of what you win. A player who is genuinely break-even at one stake often turns a modest profit a level higher for this reason alone, provided the opposition doesn’t stiffen faster than the rake lightens. At the very bottom of the online ladder that trade is often favorable. At the step from mid to high stakes it almost never is.
The other thing that changes is what a mistake costs in your head rather than in blinds. Stack sizes are identical in big blinds. They are not identical in dollars, and your brain is not a solver.
A shot-taking structure that lets you fail cheaply
Don’t move up. Take shots, with the terms written down before you sit.
Worked example. You have $2,500 playing $0.50/$1, where a buy-in is $100. That’s 25 buy-ins. At $1/$2 the same roll is 12.5 buy-ins, which is thin. Rather than wait to double your bankroll, carve out a shot fund: three buy-ins, $600. You play $1/$2 until either the $600 is gone or you’ve added $600 in profit at the new level.
If the fund dies, you go back down. Not for a session. Back down until you’ve rebuilt to $2,800, which means winning roughly six buy-ins at the lower stake. That rebuild requirement is the part everyone skips, and skipping it is how a shot becomes a permanent demotion to a stake you can’t afford.
If the shot wins, you don’t automatically stay. You’ve got $3,100 and a 300-hand sample, which proves nothing. Take a second shot on the same terms. Three successful shots and a bankroll past 25 buy-ins at the new stake is a move up. One good session is a good session.
Two conditions should void a shot entirely. The first is a game you haven’t observed. Sit out for fifteen minutes and watch, or at a live table, take a walk past it. The second is your own state: a shot taken on a day when you’re already irritated at the level below isn’t a shot, it’s a reaction. The quiet leaks that drain winning players show up at a new stake first, because that’s where your defaults are weakest.
When the usual advice is wrong
Three cases. First, the recreational player with income. If you earn $8,000 a month outside poker and you’re playing a $300 buy-in game for entertainment, bankroll rules built around risk of ruin don’t bind you. Your real constraint is whether you’ll play scared, and whether losing four buy-ins ruins your week.
Second, game selection beats stake selection. A soft $2/$5 table with three recreational players is a better place to be than a $1/$2 table of regulars, even if your bankroll says otherwise on paper. Variance is the price; the edge is bigger. Take a shot at the soft game and be honest when it breaks up.
Third, the games above you may not exist in a year. Online player pools have been thinning at the middle and upper stakes for over a decade, and live room schedules follow local demand. Waiting for perfect proof of a win rate you’ll never measure, while the game you were saving for disappears, is its own kind of error.
The strongest signal that you’re ready isn’t a number in an account. It’s that you’ve stopped being surprised. You know what your opponents do, you know what you’d do against each of them, and the hands you lose are hands you’d play the same way tomorrow. When that’s true at your current stake, the next one is a question of logistics.