The noise is larger than the edge
A session is a tiny sample and a hand is smaller still, so most of what happens at a table is the cards rather than the player. The uncomfortable part is that this cuts both ways: a bad run is not proof of bad play, and a good one is not proof of skill.
Hand 01What variance actually is here
Variance is the spread of outcomes around their average. In poker the spread is enormous for a simple reason: the outcome of a hand is decided by cards that are hidden until they are not, and a single card can turn a large win into a large loss. The average result of a well-played decision may be positive while any individual instance of it is negative.
This is not an excuse for poor play, and it is not a comfort. It is a description of the measurement problem: when the spread of outcomes is much larger than the average edge, a small number of results cannot distinguish a skilled player from a lucky one. Confusing the two directions of that error is why players mislearn from their own experience more often than from anything they read.
- A hand
- One instance of a decision. Almost pure noise about quality.
- A session
- A few hundred decisions at most, and often fewer. Still dominated by noise.
- A sample
- A number of hands large enough for the average to show through. Larger than most players expect.
- A fee
- A certain, repeated cost that makes the average worse and the sample needed longer.
Hand 02The error runs in both directions
A losing run tempts a player into changing a process that was fine, and a winning run tempts the same player into concluding that a process was good when it was merely fortunate. Both conclusions come from the same mistake: reading a small sample as though it were evidence about quality.
The practical remedy is to judge decisions rather than results, and to judge them against what was known at the time rather than what turned out. A call that was correct given the information available is correct even when the river card was unkind, and a call that was wrong is wrong even when it happened to win. See pot odds for how that judgement is made explicit.
Normal and expected. Nothing to change, and nothing to celebrate.
The dangerous one. A reward that teaches exactly the wrong lesson.
What everyone hopes confirms their judgement, and it does not by itself.
Painful, and easy to over-correct from — the fix may be the opposite of the problem.
Hand 03Why the fee makes the sample problem worse
A rake takes value out of every pot regardless of the result, so the average outcome of the whole table is negative. That does not add variance directly; it shifts the average downward, which means a player must win by more before any result can be attributed to skill rather than to noise.
It also changes what a win looks like. A session that finishes slightly ahead at a small stake may be a losing session once the fees paid during it are counted, because the fee is charged per pot rather than at the end. Reading results without accounting for what the table charged is the easiest way to convince yourself that a losing process is a winning one. See the rake for how the fee scales with stake.
Hand 04What a proper sample would require
A sample large enough to outweigh the noise is much larger than a session, a week or often a running month, and it is larger at low stakes where the fee is proportionally heavier. That is not a reason for despair; it is a reason to stop using results as a scoreboard and start using decisions as one.
The record that actually helps is a record of process: what was known, what was decided, and whether the decision would be made again with the same information. That record can be built in a single session, because it does not depend on the cards at all, which is precisely why it is more useful than a graph of results over the same period.
Keeping a usable record
- Record the decision and the information available, not the outcome.
- Count the fee paid, so the session total reflects what the table charged.
- Review a period, not an evening, when judging whether anything has changed.
- Treat any conclusion drawn from a handful of results as a hypothesis rather than a finding.
Why the game is a transfer with a fee removed
Players are paid by other players, minus what the room keeps. No amount of skill removes that; it is the arithmetic of the product rather than a gap that can be closed.
Affiliate disclosure and risk warning
Every affiliate link on this page and in the header is a sponsored link to a partner operator, and we may be paid if you open an account through it, at no extra cost to you. That link pays us; it does not make any game cheaper to play, it does not improve any decision, and it is never a recommendation to play. Nothing on this page is legal, financial or poker advice. 18+ only. Gambling is a real risk of real loss. Every cash game described here takes a fee out of every pot it deals and every tournament keeps a share of every buy-in, so the average player is behind before a single card is dealt, and most players lose money over any meaningful number of hands. No strategy on this site changes that arithmetic; it is the funding model of the game, not a skill gap. Whether poker may be offered where you are, who may play, whether the operator is licensed, what is owed on winnings and whether any arrangement between players is lawful all differ between countries, states and provinces, and they change; a general explanation of how these games are built is not advice on your own position or your own eligibility. Never stake money you cannot afford to lose, never borrow to play, and never chase losses with a larger buy-in. Results are driven by cards nobody controls, and a short run of results tells you almost nothing about the quality of your play. Gambling can cause serious financial harm, including debt and damage to relationships and mental health. Free and confidential support is available in most countries through national gambling-harm helplines, for players and for the people around them.