Open problem, Combinatorics and discrete mathematics, Geometry, posed 1935
Happy ending problem (Erdős–Szekeres)
OpenErdős
For every integer , does every set of points in the Euclidean plane in general position (no three collinear) contain points that form the vertices of a convex -gon—so that the minimum guaranteed number of points is ?
As of 2026, the exact formula is proved only for , with , , , and ; even the case (conjectured to be ) remains open. Asymptotically, following Andrew Suk's 2016 breakthrough , Holmsen, Mojarrad, Pach, and Tardos (2020) improved the upper bound error term to , leaving only a sub-exponential factor between the lower and upper bounds.
Best known results
- Exact values: , (Klein, 1933), (Makai–Turán, 1935), and (Szekeres–Peters, 2006).
- For all , (Erdős–Szekeres 1960; Suk 2016/2017; Holmsen–Mojarrad–Pach–Tardos 2020).
Tools and where they stop
| Tool | Achieved | Where it stops |
|---|---|---|
| Positive-fraction cups-and-caps decomposition (Pór–Valtr, Suk) | Uses the Pór–Valtr positive-fraction theorem to partition most of a point set into convexly independent subsets whose points lie in narrow cones, enabling simultaneous induction on cups and caps to lower the base from to . | Passing to a positive fraction at each recursive depth loses a factor of per stage, accumulating the overhead above . |
Open questions
- Does every set of points in the plane in general position contain the vertices of a convex heptagon ()?
- Is for every integer ?
References
- Paul Erdős, George Szekeres (1935). A combinatorial problem in geometry
- George Szekeres, Lindsay Peters (2006). Computer solution to the 17-point Erdős-Szekeres problem · DOI:10.1017/S144618110000300X
- Andrew Suk (2017). On the Erdős-Szekeres convex polygon problem · DOI:10.1090/jams/869 · arXiv:1604.08657