How Long Poker Downswings Last and How to Survive Them

Break-even stretches of 20,000 to 30,000 hands are routine for a solid winner, and a 5bb/100 player can sit below a previous peak for 100,000 hands with nothing wrong in their game. Depth matters more than duration: 20 buy-ins under your high-water mark is ordinary, 30 is unpleasant and still well inside normal. The real question is never whether the downswing is real. It is whether your bankroll and your decisions can absorb it while the maths catches up.

What a downswing actually is

A downswing is the distance between your current balance and your highest previous balance. It is not the distance below zero, and it is not the feeling that the deck has turned against you.

That distinction splits downswings into two types, and only one of them looks dramatic.

Losing stretches are the obvious kind. You lose flips, you get coolered, the graph points down at an angle.

Break-even stretches are worse. Nothing memorable happens. You play 40,000 hands and finish exactly where you started, having expected to win 20 buy-ins. Your bankroll is unharmed and you have quietly lost two months of income.

Most long downswings are the second kind. Players brace for a cliff and get a plateau instead, which is harder to sit through because there is no single hand to blame.

How long they last, measured in hands

Time is the wrong unit. A downswing has a length in hands, and your volume converts that into weeks or years.

• Online, four tables: roughly 300 to 350 hands an hour, so 100,000 hands is about seven weeks of full-time play.
• Online, one table: roughly 80 hands an hour, and the same stretch takes most of a year.
• Live: roughly 25 to 30 hands an hour, which makes 100,000 hands a three-year project.

This is why live players and online grinders describe completely different worlds. A live player's miserable year is 15,000 hands, a sample so small it says almost nothing. An online grinder runs through that in a fortnight and knows it means nothing.

For a winning 6-max cash player, expect break-even or losing runs of 20,000 to 30,000 hands several times a year. Runs of 50,000 hands below a peak happen. Runs of 100,000 hands happen to small winners.

Why the maths guarantees it

Two numbers describe your results: your win rate in big blinds per 100 hands, and your standard deviation in the same units. For 6-max no-limit cash that second number usually sits near 90, higher if you play a loose aggressive style and lower at full ring.

Here is the whole problem in one line.

Your edge grows in proportion to hands played. Your swing grows with the square root of hands played. Downswings end because the first term outruns the second, and they last a long time because the square root is patient.

Over 100 hands, a 5bb/100 winner expects 5bb against a standard deviation of 90bb, so the noise is 18 times the signal. Over 100,000 hands the expectation is 5,000bb against a standard deviation of about 2,850bb, so the noise is a little over half the signal. Better, and still nowhere near settled.

That relationship produces the table every serious player should see once:

Hands played	95% range of observed win rate	Chance of showing a loss
10,000	-12.6 to +22.6	about 29%
25,000	-6.2 to +16.2	about 19%
50,000	-2.9 to +12.9	about 11%
100,000	-0.6 to +10.6	about 4%
500,000	+2.5 to +7.5	under 1%

Every row assumes a player whose true win rate is 5bb/100 with a standard deviation of 90. A genuine winner looks like a losing player on paper roughly one time in three after 10,000 hands. The mechanics behind that spread are unpacked in standard deviation in poker.

How deep the hole gets

Depth is what actually threatens you, because depth is what empties a bankroll.

Play enough hands at 5bb/100 and a 10 buy-in drawdown becomes near certain, a 20 buy-in drawdown becomes likely, and a 30 buy-in drawdown stays uncommon without ever being rare. Two relationships govern all of it.

Halve your win rate and every drawdown depth roughly doubles. A 2.5bb/100 winner takes the swings of a 5bb/100 winner at twice the scale.

Double your standard deviation and depths grow fourfold. This is the hidden cost of a high-variance style, and the reason two players with identical win rates can have wildly different experiences of the same month.

Those relationships are the entire justification for bankroll rules. The usual 6-max cash guideline of 30 to 50 buy-ins is not superstition, it is a number chosen so a normal 20 to 30 buy-in drawdown cannot end your career. Set your bankroll cushion for downswings while you are winning, because the number you pick mid-downswing will always be too small.

Rather than trusting general figures, feed your own win rate, standard deviation and volume into a downswing simulator and look at the worst run across a few thousand samples. Watching your own inputs produce a 25 buy-in hole is the fastest cure for believing yours is unusual.

Downswing or leak?

Variance and a leak feel identical from the inside. In a database they look nothing alike.

Variance is scattered. Sound decisions produce bad outcomes across unrelated spots: set over set, flush over flush, a run of flips lost. No pattern by position, stack depth or line.

A leak is a category. One filter loses money every time you apply it — your BB defence against BTN opens, your turn barrels on paired boards, your 3-bet pots out of position.

Three checks separate them in about an hour.

1. Compare all-in EV to actual winnings. A wide gap with a healthy EV line means you ran badly. If the EV line is flat too, you are not running badly, you are playing badly.
2. Filter by position. Losing 8bb/100 from the BB is normal and expected. Losing 30bb/100 from the BB is a defence problem.
3. Split showdown and non-showdown winnings. A collapsing non-showdown line usually means you have started giving up too often, which is a response to a downswing rather than a cause of one.

The trap is that checks two and three often go bad because of check one. You take a genuine bad beat, tighten up, stop bluffing, and convert a run of bad luck into a real drop in win rate. That is the danger of a long downswing: not the money variance takes, but the money your adjustments give away on top.

Surviving the middle

Move down on a number, not a feeling. Decide in advance the bankroll level at which you drop a stake and obey it without negotiation. Moving down at 25 buy-ins and back up at 35 removes the decision from the moment you are least able to make it.

Cut session length before you cut volume. Fatigue and losing compound each other. Four two-hour sessions beat one eight-hour session in any month, and the gap widens when the graph is going the wrong way.

Set a stop-loss in buy-ins. Three buy-ins down is a reasonable trigger to close the tables. It is not a strategic rule, it is a guard against tilt after a brutal session bleeding into the next one.

Review the spots that cost you, not the spots that hurt. The hand that busted you is rarely the hand that lost you money across 30,000 hands. Set a filter, not a grudge.

Put volume on the whiteboard instead of profit. Hands played is the only target you fully control, and it is the only one that shortens the downswing.

When it is worth worrying

Four signals justify a genuine review rather than patience.

• Your results are negative across 100,000-plus hands at the same stake against the same pool. That is roughly 4% territory for a 5bb/100 winner, so the simpler explanation is that you are not one.
• Your all-in EV line matches your winnings and both are flat. Nothing is being taken from you by luck; the edge is not there.
• The pool has changed. Regulars have adjusted, the recreational traffic has moved, and the strategy that beat the game last year is now the game.
• You cannot play your normal game any more. A downswing you are handling badly is a real problem even when the maths says it is ordinary.

Everything else is waiting. The uncomfortable truth about downswings is that the correct response to most of them is to keep playing well, keep the sessions short, and let a very slow-moving piece of arithmetic do its work.