ChessPivot
Young woman at a table, absorbed in her game of chess

One More Game

Does chess addiction exist?
Alex Engelman
Contents

You told yourself one game. It is one in the morning and you have played eleven. You have already deleted the app, and you have already reinstalled it. Every chess community has a name for this scene: they call it an addiction.

The word actually hides two different questions, and they do not have the same answer. The first is medical: is there an illness here? The second is practical: is something genuinely making it hard to stop? Collapse the two and you end up either treating a passion as a symptom, or waving away what thousands of people plainly describe.

So we took them one at a time. This article keeps only what rests on data: the scientific literature where it exists — and the emptiness, where it does not — plus our own measurements across two hundred and sixty-five thousand games. Every figure points to its source, and all of them are free to read.

The essentials

If you read only one block of this article, read this one.

A word nobody had measured

Online chess has changed scale. Chess.com announced two hundred and fifty million accounts in February 2026 (Chess.com, February 2026), and Lichess, which publishes all its data, recorded a little over eighty-nine million rated games in July 2026 alone (Lichess open database) — close to 2.9 million a day.

With that scale came a vocabulary. So we went looking for what research says about it, across the major scientific and medical archives — roughly forty million published papers. What we found was a void: no survey of how common it is, no questionnaire built to measure it, no published description of a patient. The only work pairing chess with addiction studies chess as a treatment — supporting recovery from alcohol dependence (Seeger et al., 2026), or as a therapeutic tool with adolescents (Gerhardt et al., 2026).

Nobody has ever measured, in online players, any of the basic things you would want to know: how many games people string together, how long a sitting lasts, what happens after a defeat, how much of it happens at night.

The official classifications do not cover it either. The World Health Organization does recognise a gaming disorder, but its definition is written around video games (WHO, gaming disorder). Should it be extended to non-digital games? The question was raised for the first time in 2026 (Sanders et al., 2026), and it remains open.

Four common arguments, and what measurement says

That leaves the reasoning everybody does. Four arguments deserve to be taken seriously, and all four crumble on contact with the data.

The most common one — the more you play, the worse you feel — has been tested properly twice, using records of what people actually played rather than what they remembered. Both times it failed: one study found that those who played most reported feeling slightly better (Johannes et al., 2021); the other, across nearly thirty-nine thousand players at seven publishers, found no evidence that playing time causes anything at all (Vuorre et al., 2022).

The second argument is the most insidious, because it sounds like common sense: he has lost interest in everything else. That is precisely the criterion specialists rate as least reliable, and one study showed that the usual risk factors for addiction do not separate struggling players from those who are perfectly fine (Deleuze et al., 2017). The criterion identifies people who play a lot, not people who are ill.

One thing does hold, though: sleep. Pooling thirty-four studies and nearly fifty-two thousand people, heavy gaming is reliably associated with shorter sleep (Kristensen et al., 2021). Modestly, but consistently. If you are going to watch one indicator, watch the hour you stop, not the number of games.

That leaves a fairer way of thinking about all this, and it comes not from research on addiction but from research on passion (Vallerand et al., 2003). It distinguishes two ways of loving an activity intensely. In the first, you choose it and it fits inside your life. In the second, it presses on you and crowds the rest out. The difference does not show up in the number of hours: it shows up in whether you are the one deciding. That distinction survives the data. The word "addiction" does not.

What people say, and what has been measured
"The more you play, the worse you feel"Measured on real play records, the link is absent — and sometimes reversed. Time spent does not predict feeling bad. (38,935 players, seven publishers)
"He has lost interest in everything else, that’s the sign"That is precisely the criterion specialists rate as LEAST reliable. It is also the first thing anyone says about any enthusiast. (29 specialists, three rounds)
"Studies say X% of players have a problem"From under 1 in 100 to more than 3 in 100, depending on the questionnaire. Most of the gap comes from the tool, not the players. (18,932 people)
"A defeat tilts you and ruins the next game"The only large-scale test finds no effect holding across players generally: a few individuals, not a rule. (Analysis on Lichess data)
None of these studies is about chess, except the last. This is stated each time: with no work on chess players available, we transpose from video games — and a transposition is not a proof. The high figure in the third row comes from a different review (Stevens et al., 2021), whose estimates were later the subject of a published correction: that is why we give an order of magnitude rather than a percentage.

The pull, though, can be measured

Since nobody had done it, we measured it. We took every timestamp from 217,797 rated games played by 302 players across all time controls, plus the play time shown on the public profiles of 298 further accounts.

The first result fits in one figure: 21.8% of games begin less than ten seconds after the previous one ended, and 48.3% within a minute. The typical gap between two games is 75 seconds. Put another way, more than one game in five starts before you could have finished asking yourself whether to play it.

And the shorter the time control, the less deliberation there is. After a bullet game — one minute each — a third of the transitions happen in under ten seconds. After a blitz game, a fifth. After a rapid game, one in ten.

One qualification that matters: all of this is measured on Lichess, a non-profit with no advertising, no daily streak to keep alive and nothing to sell you. We chose it because it is the only platform that publishes its data, not because it is the most manipulative. Elsewhere, these figures are not lower.

Games restarted in under ten seconds
210,723 game-to-game transitions, tournaments excluded · share of next games started within ten seconds
After a bullet game34,2 %After a blitz game21,5 %After a rapid game10,5 %0 %40 %
Our measurements, September 2026. Tournament games are excluded: pairing there is automatic, so the player does not choose to play again.

After a defeat, people stop more often — and restart faster

The next result has two faces that seem to contradict each other, and that is exactly what makes it interesting.

A defeat more often pushes people to close: 34.6% of lost games are the last of the sitting, against 29.5% of games won. But when the player does not close, the defeat sends them straight back in: 37.7% of restarts happen within ten seconds after a loss, against 27.3% after a win.

Deciding whether to continue and deciding how fast are therefore two different things, and a defeat pushes them in opposite directions. The draw, for its part, is the most exasperating result of all: nothing produces more instant restarts.

What the player does immediately afterwards
206,678 games, tournaments and disconnections excluded · a sitting = games less than thirty minutes apart
Ends the sittingRestarts within ten seconds
After a win29,5 %27,3 %After a draw30,7 %41,2 %After a defeat34,6 %37,7 %0 %45 %
Games ending in a disconnection are excluded: losing your connection produces both a defeat and the end of a sitting, which would manufacture the very result we are trying to measure.

And four players in ten do the opposite

That average hides two populations. Looking player by player — each serving as their own control — most do indeed stop more often after a defeat. But of the 293 players with enough games to be judged, 116 do the opposite: they play more after losing.

So the useful question was never "do chess players chase their losses?" Most do not. It is: which ones do? And the answer sits at around four in ten.

What a sitting actually looks like

Here is the reassuring figure, and it is an honest one: the typical sitting is two games and twelve minutes. Most of the time, people do not play much.

But almost everyone occasionally does far worse. Of our 302 players, 253 have had at least one sitting of ten games or more, half have had one of twenty, and the longest run we found is 145 games back to back. Above all, those long sittings matter far more than their rarity suggests: they are only 11% of sittings, but 29% of all games played.

The marathon is not daily life, then. Nor is it a curiosity.

What a sitting contains
69,299 sittings reconstructed across 302 players · a sitting = games less than thirty minutes apart
The median sitting212 min
9 sittings in 10 under754 min
1 sitting in 100 exceeds222 hr 35
The longest observed14510 hr 20
Our measurements, September 2026. Sittings of ten games or more are only 11% of sittings, but 29% of the games played.

A correction we owe you

We nearly published something false here, and you deserve to know what. Sorting games by their rank within the sitting, playing quality clearly declined: by the seventeenth game, players were performing noticeably below their own average. The headline was ready — the fifteenth game costs you.

Then we realised that two explanations produce exactly the same curve. Perhaps long sittings wear you down. Or perhaps people simply stop after playing badly. So we realigned the sittings on their last game instead of their first.

The decline sits entirely in the final game. Remove it, and it vanishes. We therefore cannot tell you that playing at length degrades your game: our own data does not show it. What we can tell you is that sittings end on a bad game. People do not stop when they are tired. They stop when they have just played badly.

Why everyone thinks everyone is addicted

Lichess shows, for every account, the total time spent playing. Among the opponents our players actually met in open matchmaking, the median value is 581 hours of chess. The figure is true, it is frightening, and it is exactly the kind of figure that circulates. It is also deeply misleading, for a reason worth understanding.

You can only be paired with someone who is playing. Someone who plays six hours a day is hundreds of times more likely to turn up across the board from you than someone who plays twice a month. A sample of opponents is therefore not a sample of players: it tilts massively towards the heaviest users. Correct for that bias and the picture changes entirely.

The average player plays very little; the average opponent plays enormously. Both figures are accurate — they simply answer two different questions. That is why the site you experience feels full of people who never stop: the people who never stop are precisely the ones filling the games. And it is why almost every alarming statistic about playing time is inflated.

Two true figures, two different questions
298 Lichess accounts met in open matchmaking · play time shown on the public profile
Median play time581 hr50 hr
At least 1 hr a day20.1%2.0%
At least 2 hr a day3.4%0.2%
The right-hand column reweights the sample by the inverse of activity, to approximate a population of players rather than of games. It is an approximation: only a study drawing players at random from the open database would give the exact figure.

What players advise each other

We read 22,026 comments from players discussing the subject with each other, drawn from 150 conversations across chess and digital-sobriety communities. Of that total, 1,905 actually proposed a solution.

The result is clear, and it does not go the way you would expect at all. Around 95% of the advice is about changing what you do with chess. Around 3% is about doing less of it. Almost nobody recommends setting yourself a daily limit.

And the community does not merely say it more often: it approves of it more. Comments recommending you join a club are voted well above average, while those recommending a blocker or deleting your account fall well below.

What players who propose something actually propose
1,905 comments carrying a remedy, out of 22,026 read · one comment may propose several
Study instead of playing49 %Play at a club or tournament46 %Do something else4 %Move to slower games3 %Install a blocker1 %Delete the app1 %Set a daily limit1 %0 %55 %
Keyword detection across 22,026 comments: the shares are indicative, but a gap of this order does not depend on the fineness of the coding.

Study rather than play

Players worked this out on their own. It also happens to be what research shows, by two entirely separate routes.

On one side, restricting yourself works badly. The subject has been studied for years in gambling, where the stakes are far heavier and the tools far more developed, and the results there are consistently disappointing. On the other, a study published this year followed 44,213 chess players over time, using their real activity records rather than their own accounts of themselves (Southwick et al., 2026). It finds that an hour spent working on your chess is worth roughly 3.6 hours spent playing games.

Note carefully what that says, and what it does not. Playing does teach you something: it is not wasted time. But the same hour spent studying would have taught you three and a half times more. If you are grinding out games because you want to improve, you have picked the slow route.

The pop-up warningA real effect while it is on screen, and nothing beyond.No lasting effect
The forced one-hour breakThe share who stop for the day rises from 27% to 68%. By the next morning, nothing remains.One day only
Voluntary self-exclusionAbout one player in sixty ever uses it. The tool exists and stays on the shelf.Almost nobody
The limit you set yourselfSix in ten who set one report going to the maximum allowed every time: the ceiling becomes a target.Backfires
Cutting screen time, all methodsAcross 204 pooled studies, the average effect is small.Weak
Changing how the tool worksAdding friction to the phone beats asking people to set themselves goals.Beats willpower
Studying instead of playingAn hour of study teaches roughly 3.6 times more than an hour of games, across 44,213 chess players followed over time.Measured, and on chess players
The first six rows come from gambling or screen-time research, not from chess: they transpose, they do not prove. Only the last one is about chess players.

What holds you is not the game, it is the counter

An earlier piece of research explains why the games are so hard to stop. Examining 133 million online games and 284,000 personal bests, its authors found that at the moment a player sets a new rating record, their probability of stopping play jumps (Anderson and Green, 2018).

In other words, people do not play chess until they are satisfied with their chess. They play until the number is where they want it. What holds you is the counter.

That is exactly what makes the advice "study rather than play" useful, and not merely virtuous. It does not ask you to love chess less, an instruction nobody ever follows. It does not treat a passion as a symptom. And it acts on the mechanism actually identified: if it is the rating that holds you, the way out is to make progress where the rating is not watching.

What we cannot tell you

What is missing must be said too, and a great deal is missing.

We measured behaviour, never suffering. None of this marks anyone out as having a problem, and the link between hours played and actually feeling bad is precisely what the best research finds weak. A high game counter is not a diagnosis, including your own.

Our timestamps come from a single platform, the one that publishes its data. Our playing-quality measurements come from our own users, who came to us because they wanted to improve: they are not a representative sample of anybody.

We also tried to check the 3.6 ratio ourselves, and failed. Across 186,480 blitz games, months in which people played more did not return less rating — rather slightly more, well within the margins of chance. And players who pack their games together gain more than those who spread them out, which looks like the opposite of our argument. But it cannot be read that way: spreading eight hundred games over three years means spending three years getting rusty between them, and those playing eight games a day are often in the middle of an improving run. Playing intensely and improving go together; which causes which, our data does not say.

Finally, the real gap remains wide open: nobody has ever taken the same chess players and crossed the record of what they did with a serious measure of how they feel. That is the study that overturned the picture for video games. Until someone does it for chess, the question "who is genuinely struggling?" has no honest answer.

In your games

Three moves, drawn from what has actually been measured. First, change the ground rather than your willpower: since it is the rating that holds you, replace the next game with twenty minutes of exercises, or with a review of the game you have just played. It is the only substitution whose return has been quantified, and that return is 3.6 to 1.

Second, be wary of the restart within ten seconds: it is the least considered decision of your whole sitting, and it is twice as frequent in bullet as in rapid. Moving from a one-minute time control to a ten-minute one mechanically changes how many times you have to decide at all.

Finally, one simple signal to watch — the one the literature establishes best: the hour at which you stop, not the number of games. The only robust link between heavy play and health runs through sleep.

Going further: the studies cited

Every figure in this article points to the study that establishes it: the call in brackets in the text leads to the matching entry below. All links open a version you can read in full for free, with one exception, flagged where it appears. Our own measurements, as of September 2026, rest on 217,797 rated games from 302 Lichess players across all time controls, on the public profiles of 298 further accounts, on 47,070 games analysed by the ChessPivot engine, and on 22,026 player comments.

Southwick, Harwell, Wright, Olsen and Ogles (2026). Not all practice is created equal: longitudinal evidence from over 40,000 chess players. Psychological Science.Full text
Anderson and Green (2018). Personal bests as reference points: 133 million online games and 284,000 personal bests. Proceedings of the National Academy of Sciences.Full text
World Health Organization. Addictive behaviours: gaming disorder. The official definition, its three criteria and the twelve-month requirement.WHO page
Sanders, Stevens, Radünz, Delfabbro and King (2026). Should gaming disorder include non-digital gaming activities? Journal of Behavioral Addictions.Full text
Johannes, Vuorre and Przybylski (2021). Video game play is positively correlated with well-being, measured on records supplied by the publishers. Royal Society Open Science.Full text
Vuorre, Johannes, Magnusson and Przybylski (2022). Time spent playing video games is unlikely to impact well-being: 38,935 players at seven publishers. Royal Society Open Science.Full text
Castro-Calvo, King, Stein, Brand and colleagues (2021). Twenty-nine international specialists rate the criteria for gaming disorder over three rounds. Addiction.Full text
Deleuze, Nuyens, Rochat, Rothen, Maurage and Billieux (2017). Established risk factors for addiction fail to discriminate healthy gamers from those endorsing the criteria. Journal of Behavioral Addictions.Full text
Przybylski, Weinstein and Murayama (2017). Four surveys, 18,932 people: between 0.3% and 1.0% might qualify for a diagnosis, and the links to health outcomes are mixed. American Journal of Psychiatry.Accepted manuscript, Cardiff University repository
Stevens, Dorstyn, Delfabbro and King (2021). Global prevalence of gaming disorder: a systematic review. This is the only reference in this article whose full text is not free, and its prevalence estimates were the subject of a correction published in 2023 — which is why we quote no percentage from it. Australian and New Zealand Journal of Psychiatry.Summary only
Gee, Seese, Curley and Ward (2025). The effect of one result on the next game, on Lichess data: no effect holding across players generally. Preprint, not yet peer-reviewed.Full text
Kristensen, Pallesen, King, Hysing and Erevik (2021). Problematic gaming and sleep: thirty-four studies, 51,901 people. Frontiers in Psychiatry.Full text
Vallerand and colleagues (2003). Les passions de l’âme: on obsessive and harmonious passion. The framework separating a passion you choose from one that imposes itself. Journal of Personality and Social Psychology.Full text
Seeger and colleagues (2026). Chess-based cognitive remediation in alcohol use disorder: modest gains in attention, no effect on abstinence. Neuropsychological Rehabilitation.Full text
Gerhardt and colleagues (2026). Chess as a therapeutic approach in adolescents: a pilot study with thirty-three participants. Clinical Child Psychology and Psychiatry.Full text
Bjørseth and colleagues (2021). Pop-up messages in gambling: eighteen pooled studies, a real effect with no lasting hold. Frontiers in Psychiatry.Full text
Auer and Griffiths (2023). The effect of a mandatory one-hour break among 2,021 online casino players: clear on the day, nil the day after. Journal of Gambling Studies.Full text
Hopfgartner, Auer, Griffiths and Helic (2023). Of 25,720 players, 414 — about one in sixty — ever used self-exclusion. Journal of Gambling Studies.Full text
Auer and Griffiths (2023). Attitudes towards deposit limits: 60.5% of players who set one report going to the maximum allowed. Journal of Gambling Studies.Full text
Jones, Armstrong, Weaver and colleagues (2021). Reducing children’s screen time: 204 pooled studies, small average effect. International Journal of Behavioral Nutrition and Physical Activity.Full text
Zimmermann and Sobolev. Design friction versus goal setting: a randomised trial with 112 participants. The link leads to the authors' preprint, under a slightly different title — the version published in Cyberpsychology, Behavior and Social Networking (2023) is paywalled.Preprint
Lichess. The database of every rated game on the platform, released into the public domain. It is the source of our counts, and of anyone else’s who wants to check them.Download the database
Chess.com (27 February 2026). The platform’s own announcement of 250 million members. This is a count of accounts, not a measure of how many people actually play.The announcement