Colophon
About the data
Problems and solutions
All 702 problems come from the official problems-and-solutions collections published by the competitions themselves – for the IMC, one PDF covering 1994–2025; for the ICPC, the per-problem PDFs the organisers post for each World Finals. Typesetting changes substantially from year to year: papers were set in different LaTeX classes, one year is imposed as a two-up booklet, another is rotated to landscape, and several embed fonts with no usable character map.
Everything is therefore captured twice. Each statement and each solution is cropped straight from the original page at 2.5× scale, with the paper made transparent and the ink black so the same image serves the light and dark themes. That crop is the reference: it cannot drift from what the organisers printed.
On top of that, 701 of the 702 statements – every edition from 1991 on – have been transcribed into LaTeX by hand and are typeset with KaTeX. Those pages give you selectable, copyable, screen-readable mathematics that reflows on a phone. Every transcription was written against the crop of the original, and every formula is compiled by KaTeX before the site builds, so an expression that does not typeset stops the build rather than reaching the page.
Solutions still render as crops, as does the one statement whose figure no transcript can replace. A flattened text layer is also extracted from the PDF for search and previews; it keeps the original Unicode symbols but loses two-dimensional layout, so a displayed fraction collapses into separate lines. It is good enough to find a problem and never used to present one.
Known gaps
- The IMC 2014 paper embeds a Type 3 font that carries no encoding, so its text extracts as shredded fragments. Its 10 problems are searchable only through their hand-written transcripts; the one that has none is excluded from keyword search.
- Solutions are not transcribed yet. They are roughly five times the volume of the statements; until they are done, revealing a solution shows the crop from the original paper.
- ICPC coverage varies by year.Every World Finals from 1991 to 2025 is transcribed from the organisers’ own problem sets; the finals before 1991 are scans of the printed papers, too rough to transcribe, so each edition links its official PDF. Team standings are mirrored for 1999–2025 from the Competitive Programming Hall of Fame – 1999–2005 list only the medal-winning field, later years the full one – and are unavailable earlier.
- IMC 2020–2022 each ran four problems per paper rather than five; the archive is complete for those years despite the smaller count.
- Points are shown only where the paper prints them. Each competition weights its own papers: early IMC editions scored problems individually (13, 15, 18, 20, 25 points), and from 2009 onwards every problem is worth 10.
Results
IMC standings are mirrored from imc-math.org.uk. Full individual and team tables are available for 2015–2025, which is as far back as the organisers publish them as web pages. For 1994–1997 and 2005–2013 the official results exist only as PDFs or spreadsheet exports; those are linked from the relevant year rather than transcribed, because re-typing thousands of names invites errors that would be impossible to spot. The remaining years have no results published at all.
ICPC team standings come instead from the community-run Competitive Programming Hall of Fame, which is where the World Finals results are collected in a form that can be mirrored. Because the ICPC ranks teams, a contestant who sat both competitions shows their IMC papers marked problem by problem and their ICPC finals as team placings, side by side in one profile.
The results page searches those scoreboards across every year by contestant or university name. Be aware that the organisers spell institutions inconsistently between editions – Eötvös Loránd University appears both ways round, Cambridge under five variants – so a search offers the union of the spellings it matched rather than pretending one canonical name exists.
Clicking a contestant opens their record across every edition they entered, marked problem by problem. Identity there is the name as printed: no name repeats within a single edition, so entries cannot collide inside a year, but two people sharing a name across different years would be shown as one. The university on each entry is the thing to check.