Does Sudoku Increase IQ? Why No Chart Can Turn Solving Time Into a Score

Type "sudoku iq" into a search box and the suggestions on 9 October 2026 include "sudoku iq 180", "sudoku iq 188", "sudoku iq 240", "sudoku iq 260" and "sudoku iq correlation". The pattern is familiar: a puzzle everyone knows, a number everyone wants, and a search for the chart that links them. We went looking for the evidence behind such a chart, in the biomedical literature (Europe PMC), in Crossref and OpenAlex, and on the web. We did not find a peer-reviewed table, formula or norm sample that converts a Sudoku solving time, or a Sudoku rating, into an IQ. What we did find is more interesting than a chart, and it says what Sudoku can and cannot tell you.

Is there a Sudoku IQ chart?

Not a validated one. The only concrete "Sudoku IQ" we could trace is a mobile app, Sudoku IQ+ by the developer Enigit, whose page on the AlternativeTo software directory describes a real-time IQ calculation based on speed, accuracy and difficulty. That is a game score. The page does not publish the formula, the weights, a norm group or any validation against an intelligence test, so the word "IQ" there names a number the app invented. An IQ is a position inside a reference sample of people, on a test with a published reliability; our guide to how IQ tests work sets out what that requires. Without a norm group a figure can be 188 or 88 and mean the same thing, which is nothing.

Pages that print a table of the form "solve a hard puzzle in X minutes and your IQ is Y" did not turn up in our searches either, although several sites were closed to automated checks, so we cannot rule them out. If you have a specific chart, the question to ask of it is the one we ask of every score on this site: what sample was it normed on, and where is the test-retest reliability?

Why can't a solving time stand in for a score?

Because most of what a time reflects is the puzzle. Radek Pelánek of Masaryk University analysed human solving data from three web portals, "thousands of hours of human solving over 2000 problems" in the paper's words: 731 puzzles from one portal (an average of 1,307 solvers each), 1,088 from another (131 solvers each) and detailed play records for 60 users on 15 puzzles from a third. He built a simple model of human solving and found two sources of difficulty: the complexity of the individual logical steps, and the structure of dependencies among steps. His best metric correlates 0.95 with the mean solution time of each puzzle, and a version with few Sudoku-specific details still reaches 0.88.

Read that correctly. The 0.95 is between a model's predicted difficulty of a puzzle and the crowd's average time on it. It is not a correlation between a person's time and their intelligence, and the portals recorded anonymous solvers with no demographics and no test scores. What it shows is that two people who time themselves on different puzzles are not comparable, and a fast time on an easy grid says little. Any honest score would need a fixed set of puzzles of known difficulty, which no Sudoku rating published on the open web provides.

What do the studies of puzzle-doers show?

The biggest sample is the PROTECT study, run from the University of Exeter and King's College London. In a paper on number puzzles, the category that includes Sudoku (Brooker and colleagues, International Journal of Geriatric Psychiatry, 2019;34(7):932-940), 19,078 healthy volunteers aged 50 to 93 reported how often they did number puzzles and took two online cognitive batteries covering reasoning, attention, processing speed, executive function, working memory and episodic memory. Every one of the 14 measures showed a significant main effect of puzzle frequency, with P values under 0.0004, and people who did number puzzles more than once a day had the best results on 10 core measures, though not all differences against every other group were significant. A companion paper on word puzzles, using the same 19,078 people, found that those who never did them scored worst.

The design is the limit. Puzzle frequency was self-reported, the cognitive tests were taken once, and the analysis compared six frequency groups. People with sharper thinking may choose, and keep up, puzzles for that reason. The paper's own conclusion calls for research on the type and difficulty of puzzles. You may also have seen that puzzle-doers have "brain function equivalent to ten years younger". That sentence comes from the University of Exeter press release of 16 May 2019, where it is stated for word puzzles on a grammatical-reasoning test, with "eight years younger" for short-term memory. It does not appear in either abstract, it is not about Sudoku and it is not about IQ, and the release itself says the team cannot say puzzles reduce dementia risk.

Is there any Sudoku-specific evidence on intelligence?

Very little. The only paper we found on Sudoku solving ability and intelligence is a three-page article in the International Journal of Computer Applications (Onokpasa, Bisandu and Ikoojo, 2019;178(43)) that uses one class of third-year computer science students at the University of Jos; its abstract poses the question and reports no result. A blog line that "a 2018 paper found a moderate correlation" could not be matched to any paper with that year and result. On the brain side, a systematic review of six imaging studies (five fMRI, one fNIRS) published in Frontiers in Neuroimaging on 20 April 2026 found that Sudoku solving engages frontoparietal networks, including the dorsolateral prefrontal cortex and parietal regions linked to executive control and visuospatial working memory, plus the anterior cingulate cortex. The authors say the small number and variety of studies "preclude firm conclusions regarding efficacy". Using a brain region is not the same as improving it.

The puzzle itself is easy to describe. A solved Sudoku grid is one of 6,670,903,752,021,072,936,960 valid fills, 5,472,730,538 "essentially different" grids once those that differ only by symmetry are counted as one, and a puzzle needs at least 17 given digits to have a single solution, a result proved by McGuire, Tugemann and Civario in 2014. The numerals are just labels, so no arithmetic is involved. It is a logic puzzle of the kind we meet in matrix puzzles and number sequences, and it draws on the same working memory and rule-finding that we separate from reasoning in working memory and reasoning are not the same.

Do crosswords do better than Sudoku?

One randomised trial points that way for crosswords, in a particular group. Devanand and colleagues (NEJM Evidence, published online 27 October 2022) assigned 107 adults with mild cognitive impairment to 12 weeks of intensive home-based training with web cognitive games (51 people) or web crossword puzzles (56), followed by six booster sessions, and measured outcomes at 78 weeks. On the 11-item ADAS-Cog, a 70-point scale where higher means more impairment, the games group worsened slightly (9.53 to 9.93) and the crossword group improved (9.59 to 8.61). The difference was -1.44 points (95% confidence interval -2.83 to -0.06, P = 0.04). Hippocampal volume and cortical thickness fell less with crosswords than with games.

Three cautions. The comparison group did computer games, not nothing. The people had mild cognitive impairment, not healthy brains, and the crosswords were web-based. And the confidence interval almost touches zero. A larger follow-up, COGIT-2, has a published protocol (240 people, three arms) but no results yet. Nothing in it tests Sudoku, so the fair reading is that crosswords have one small positive trial in patients and Sudoku has none.

Does puzzle practice carry over to general intelligence?

The wider brain-training literature says mostly no. A review by Simons and colleagues in Psychological Science in the Public Interest (2016) found extensive evidence that training improves the trained tasks, less for closely related tasks, and little that it improves distantly related tasks or everyday cognition. A 2019 review by Sala and Gobet in Trends in Cognitive Sciences concluded that cognitive training has a minimal effect on domain-general cognitive skills, and a second-order meta-analysis by Sala and colleagues in Collabra found that once placebo effects and publication bias were controlled, the overall far-transfer effect was zero. None of these abstracts names Sudoku, so they do not prove that Sudoku cannot help. They do set the prior: practice makes you better at the puzzle. Our piece on whether brain-training apps work goes through the same evidence.

• Enjoying Sudoku is a fine reason to do it; a higher IQ is not an established one
• Compare Sudoku times only on the same puzzle, because difficulty swamps ability
• Treat any "Sudoku IQ" without a norm group as a game score
• For an IQ-style number, use a test with a norm sample and a stated reliability

Our IQ Metrics test reports a percentile among people, and the IQ percentile calculator shows what any score means as a rank. If you like logic puzzles for their own sake, our logical reasoning test works on the same skills, and the guide to Raven's matrices scores shows how a reasoning test turns answers into a standardised number.

Common questions