Observe
It is the question every losing stretch provokes, and it has an answer. Not a certain one, but a useful one: whether a stretch like this is ordinary for a player with your win rate and your swings. For most rough stretches, it is.
Poker results are a random walk with a small drift. Over a stretch of
h hours, a player who wins w big blinds an
hour with a standard deviation of s big blinds an hour
finishes with an expected w × h and a spread of
about s × √h.
The drift grows with time. The spread grows with the square root of time. That single mismatch is the whole story of a downswing: over a weekend the spread dwarfs the drift, and over a year the drift wins.
Take a live winner at 5 big blinds an hour with a standard deviation of 50. Those two numbers are an example and nothing more, but the shape they produce is the shape everyone's looks like.
| After | Expected | 95% band | Chance of being behind |
|---|---|---|---|
| 20 hours | +100 bb | -338 to +538 bb | 32.7% (1 in 3) |
| 50 hours | +250 bb | -443 to +943 bb | 24.0% (1 in 4) |
| 100 hours | +500 bb | -480 to +1,480 bb | 15.9% (1 in 6) |
| 250 hours | +1,250 bb | -300 to +2,800 bb | 5.7% (1 in 18) |
| 500 hours | +2,500 bb | +309 to +4,691 bb | 1.3% (1 in 79) |
| 1,000 hours | +5,000 bb | +1,901 to +8,099 bb | 0.1% (1 in 1,278) |
The 95% band is where 19 stretches out of 20 finish. Anything inside it is unremarkable, which is a different claim from "good".
Read the last column again. This player is a genuine winner, and after a hundred hours at the table they are still behind one time in six. Not because anything went wrong. That is simply what 100 hours of this much variance looks like.
A hundred hours is two or three months of playing a couple of nights a week. Most players have never had a sample long enough to be confident about their own win rate, and have had many samples long enough to feel certain they are broken.
The table above is somebody else's. Two players at the same stake can have standard deviations 40% apart, and the band scales directly with it: the loose aggressive player's ordinary month would be a crisis for the nit and the nit's would be a quiet week for them.
There is no way around measuring your own. You need the result and the duration of each session, in big blinds so that stakes pool together, and you need enough of them that the estimate is not itself noise. Twenty hours and eight sessions is the floor where the numbers start to mean something, and it is a floor rather than a target.
One thing the arithmetic cannot do: prove you are playing well. A stretch inside the band is consistent with running bad and with a leak that costs you less than the noise hides. What it rules out is the panic, which is worth a great deal on its own, because the panic is what makes the leak.
One stretch in forty finishes below a 95% band, so seeing one is not evidence of much. Two things are worth doing anyway, and neither is quitting.
And the thing that is genuinely not variance: the decisions you make in the ten minutes after a hand goes wrong. Those are not in any band, and they are the only part of a downswing you control.
Log estimates both numbers from your own completed sessions, draws the band around your real curve, and answers the question at the top of this page in a sentence. A session takes under ten seconds to record. All of it is free, it works offline, and it needs no account: nothing leaves your device unless you sign in to carry it to another one.