integrity
One hand is luck. A hundred thousand is skill. The middle is the problem.
Whether poker is a game of skill has been answered differently by different courts, in different countries, under different laws, and anyone who tells you it is settled is selling something. What can be settled is the arithmetic underneath the argument, and that part is not a matter of opinion. This page gives both, with the sources, and no legal advice.
sources checked 30 August 2026
the part that is arithmetic
Start at the two ends, where nobody argues. Before the flop, a pair of aces against a pair of kings wins 81.9 percent of the time on average, counted across every one of the 1,712,304 boards the deck can still produce. The strongest possible position in a single hand still loses about one time in five, which is why one hand tells you nothing about anybody.
At the other end, over a large enough sample, the luck cancels and what is left is how you played. The only real question is how large is large enough, and that has an answer you can compute rather than debate.
A winrate is an average, so it comes with a margin of error, and the margin shrinks with the square root of the sample. In big blinds per 100 hands, the standard error is the standard deviation divided by the square root of the number of hundreds of hands you played, and the familiar 95 percent band is about 1.96 times that.
The one thing that has to be assumed is the standard deviation. Take 100 big blinds per 100 hands, which is the right order of magnitude for six handed no limit hold’em; your own figure depends on your style and your format, so treat the table as the shape of the problem rather than as your personal numbers.
| Hands | Standard error | 95 percent band |
|---|---|---|
| 1,000 | 31.6 bb/100 | plus or minus 62 bb/100 |
| 10,000 | 10.0 bb/100 | plus or minus 20 bb/100 |
| 100,000 | 3.2 bb/100 | plus or minus 6.2 bb/100 |
| 1,000,000 | 1.0 bb/100 | plus or minus 2.0 bb/100 |
Read the first row again. Over 1,000 hands, an honest measurement of your winrate comes with a band about 62 big blinds per 100 wide on either side. A strong small stakes winrate is somewhere around 5. The error bar is more than ten times the thing it is measuring, which means a thousand hands can make a good player look like a disaster and a losing player look like a genius, and both of them will believe it.
The last row is the other half of the honesty. Even at 100,000 hands the band is about 6 big blinds per 100, so a genuine 5 big blind per 100 winner still cannot be cleanly told apart from someone breaking even. That is the middle nobody advertises: long past the point where skill dominates the outcome, your own results are still a poor instrument for measuring it.
This is not an argument that skill does not exist. It is an argument that your balance is a slow and noisy way to detect it, which is why the score we build is made out of decisions rather than results, and why the leak detectors refuse to report anything until they have seen enough hands to say it.
what researchers have measured
Four findings, all from sources linked at the bottom. The first two are the ones usually quoted; the third is the one usually quoted wrongly.
Skilled players make money, at scale
Levitt and Miles looked at the 2010 World Series of Poker and compared players who had been identified as highly skilled in advance with everybody else. The first group returned over 30 percent on their entries. The second lost 15 percent. Predicting in advance who will do well is the part that makes this evidence rather than hindsight.
Results persist, which is what skill looks like in data
A study of hundreds of millions of player-hand observations from real money online ring games found that players who were in the top or bottom of profitability in one period were substantially more likely to be there again in the next. Chance does not repeat itself. Something that predicts the following period is doing so because it is a property of the player.
And it takes about 1,500 hands before skill dominates
The same paper puts a number on the crossover: their simulations point out that skill dominates chance when performance is measured over 1,500 or more hands of play. Note carefully what that sentence does and does not say. It is about when the skill component starts to outweigh the chance component in outcomes. It is not a promise that 1,500 hands will tell you your own winrate, which is a different question with a much less flattering answer.
Machines beat the best humans, which settles a related question
The programs that beat top professionals at heads up and then at six handed no limit did it by playing better, not by getting luckier over 120,000 hands. A game where a measurably better strategy wins measurably more money is a game where strategy matters. That is covered on our page about bots.
The machine results, with the papers and the numbers behind them, are on the page about poker bots.
what courts have said
Two decisions we read, and the reason neither of them is the answer.
One American court said skill, and the court above it declined to answer
In 2012 a federal district judge held that Texas Hold’em is predominated by skill and therefore fell outside the statute the defendant was charged under. In 2013 the appeals court reinstated the conviction, but it did so on the wording of the statute and stated that the question of whether skill or chance predominates in poker was inapposite to the appeal. The higher court did not overturn the skill finding. It said the finding was not what the case turned on, which is a different thing and a useful one to know when someone quotes either half at you.
A German court called it a mixture, for tax purposes
The German federal tax court held in 2015 that tournament poker, in the Texas Hold’em and Omaha variants, is generally not to be classified as pure chance but as a mixture of chance and skill. That framing was reached in an income tax case, which is exactly the point: the answer was produced by a tax question, and it does not automatically travel to a licensing or criminal one.
The question is not even one question
Whether poker is a game of skill can be asked under criminal law, under tax law and under licensing law, and one country can answer it differently in each. A ruling about whether winnings are taxable income tells you very little about whether a game may be offered commercially, and neither tells you what is legal where you live.
Which is why this page gives no legal advice
We are a training product, not a law firm, and rules on gambling move. The two decisions above are named so you can read them yourself, and where you may play is a question of licensing in your own jurisdiction rather than a question about the nature of the game.
Where you may play, and how to check whether an operator is licensed where you live, is a separate question with a checkable answer, and it is covered on the page about where you can play. If the reason you are asking is that the swings have stopped feeling like swings, start instead at responsible play.
sources
Read on 30 August 2026. Two other frequently quoted studies are deliberately absent, because we saw them only in summaries and not at the source.
- United States v. DiCristina, 726 F.3d 92 (2d Cir. 2013)
The district court had held in 2012 that poker is predominated by skill and therefore outside the statute. The appeals court reinstated the conviction on the text of the statute and said in terms that the skill question was inapposite to the appeal, leaving the finding itself undisturbed and undecided.
- Bundesfinanzhof, judgment X R 43/12 of 16 September 2015
The German federal tax court classified tournament poker, in the Texas Hold’em and Omaha variants, as a mixture of chance and skill rather than pure chance, in the context of income tax.
- Levitt and Miles, The Role of Skill Versus Luck in Poker: Evidence from the World Series of Poker, NBER working paper 17023 (2011), published in the Journal of Sports Economics (2014)
Players identified in advance as highly skilled returned over 30 percent on their entries at the 2010 World Series of Poker, against minus 15 percent for everyone else.
- Potter van Loon, van den Assem and van Dolder, Beyond Chance? The Persistence of Performance in Online Poker, PLOS ONE (2015)
Hundreds of millions of player-hand observations from real money ring games. Earlier profitability predicts later profitability, and their simulations put the point where skill dominates chance at 1,500 or more hands.
next
If results are the slow instrument, use a faster one.
A decision can be judged the moment you make it. A winrate needs a sample most players never reach. The score we build works from the first kind, and the page explaining it also says plainly what it cannot know about you yet.
questions
Is poker a game of skill or a game of chance?
Both, in proportions that depend entirely on how long you look. Over one hand chance decides almost everything: the best starting hand in poker still loses to the second best about one time in five. Over a hundred thousand hands skill decides almost everything. The uncomfortable part is the middle, which lasts longer than most players expect.
Have courts decided whether poker is a game of skill?
Different courts, in different countries, answering different legal questions, have reached different framings. An American district court held in 2012 that skill predominates, and the appeals court that reversed the outcome in 2013 expressly said the skill question was not what the appeal turned on. A German tax court held in 2015 that tournament poker is a mixture of chance and skill rather than pure chance. Neither is a general answer, and neither is legal advice.
Does research show poker is a game of skill?
The studies we read point that way. Players identified as strong in advance made over 30 percent on their entries at the 2010 World Series of Poker while everyone else lost 15 percent, and a large study of online cash game hands found that profitability persists from one period to the next, with skill dominating chance beyond roughly 1,500 hands.
How many hands before I know if I am a winning player?
More than you want. Assume a standard deviation of 100 big blinds per 100 hands, which is the right order of magnitude for six handed no limit. Over 1,000 hands the 95 percent margin of error on your measured winrate is about 62 big blinds per 100, which is more than ten times the size of a good winrate. Even over 100,000 hands the margin is around 6 big blinds per 100, wide enough that a genuine 5 big blind per 100 winner cannot yet be told apart from a break even player.
If results take that long, what should I measure instead?
Your decisions. A decision can be judged the moment it is made, against what the situation called for, while a result has to wait for a sample you may never reach. That is the whole reason our score is built on decision quality rather than on your balance.
Is poker gambling?
Where money is at stake, treat it as gambling regardless of how the skill question is answered, because that is how the law where you live is most likely to treat it and because that is the safer assumption for your own bankroll. Whether a particular site may take your money is a licensing question with a checkable answer.