Mathematicians Still Struggle with a 120-Year-Old Chessboard Puzzle

The 'no-three-in-a-line' problem, first posed in 1906, challenges mathematicians to place the maximum number of points on a grid without any three lying on a straight line of any slope. Despite efforts by researchers like Serge Belongie and David Eppstein, and even advances from artificial intelligence, a complete solution remains elusive. The problem highlights deep mathematical complexities that persist even in seemingly simple configurations.
In November 1906, Henry Dudeney posed a deceptively simple puzzle in the London Tribune: can 16 pawns be placed on a standard chessboard such that no three lie on a straight line of any slope? This question evolved into the "no-three-in-a-line problem," a longstanding challenge in discrete geometry and combinatorics. Over a century later, the problem remains unsolved, with researchers striving to determine whether an n×n grid can accommodate 2n points without any three aligned—regardless of slope. In recent months, a competitive race has emerged, with Marijn Heule of Carnegie Mellon University and Thomas Prellberg of Queen Mary University of London exchanging records: Heule achieved a 2n solution for n = 76 in August 2026, only to be surpassed by Prellberg’s n = 78 solution in October. These advances build on earlier work, including Paul Erdős’s algebraic approach using parabolas and a 1975 result that achieved 1.5n points. Achim Flammenkamp’s long-running scoreboard has tracked progress since 1992. Meanwhile, a team of six mathematicians from Oxford, Caltech, and IST Austria recently solved a generalized version of the problem for k ≥ 3, proving kn solutions exist for large n—but left the original k = 2 case, dubbed “notorious” by the researchers, unresolved. Peter Keevash noted the mathematical method fails for k = 2 due to a divergent infinite sum, suggesting the need for entirely new techniques. Richard Guy and Patrick Kelly’s 1968 conjecture that solutions peak around 1.81n and then decline has gained renewed attention, with Prellberg’s 2026 preprint suggesting a possible crossover at n ≈ 493—though he cautions it is a crude heuristic. The core challenge persists: despite AI and computational advances, no one has proven whether 2n solutions exist for arbitrarily large n.
The enduring mystery of the no-three-in-a-line problem illustrates how even well-defined mathematical puzzles can resist solution despite computational progress. While AI has solved complex problems in fields like protein folding and game theory, this case shows that combinatorial geometry remains a frontier where human intuition and algorithmic search still struggle. The problem’s resistance may reflect deeper structural limitations in current mathematical frameworks, especially in handling constraints across infinite slope classes. Its persistence could influence how researchers approach other open problems in discrete mathematics, emphasizing the need for novel theoretical tools over brute-force computation. Comparable to the long-standing difficulty of the Collatz conjecture, this puzzle may serve as a benchmark for evaluating new mathematical methods. If future breakthroughs emerge, they could inspire new approaches in algorithm design, cryptography, or data layout optimization—where avoiding collinear patterns matters. However, the lack of a solution also underscores the limits of current AI in pure mathematical discovery, especially when problems involve non-local constraints. What’s next may depend on whether a new proof technique emerges, possibly from the study of algebraic geometry or probabilistic combinatorics. The community may watch closely for any sign that the 2n conjecture is falsified or confirmed, which could reshape expectations about the boundaries of mathematical solvability.