The State of
Machine Learning Competitions

Delivering yearly insights and analysis from the ML competitions ecosystem
since 2022.

Read by AI researchers, practitioners,
and tool developers.

Widely-read, trusted research for a broad technical audience — researchers comparing modelling techniques, practitioners evaluating libraries, and developers tracking uptake of their tools.

1,400+

competitions reviewed

300+

winning solutions analysed

In-depth analysis, extensively researched.

Original research, with a mix of top-down and bottom-up data gathering approaches followed by systematic analysis. Combines data provided by competition platforms, interviews and surveys of competition winners, and winning solution source code.

Cited by academic research.

From particle physics to cancer research, previous reports have been cited by over 20 academic publications — including journal and conference papers, as well as Masters and PhD Theses in several countries.

Past Reports

Every edition analyses trends in winning solutions:
common algorithms, modelling techniques, tools and libraries,
plus compute and hardware use.

Valuable insights about machine learning,
as discovered by the ML competitions community.

Competitions are increasingly important, with millions of dollars in total prize money across hundreds of competitions each year, and a growing focus on competitions as dynamic benchmarks for evaluating frontier AI systems.

On aggregate, ML competitions provide a powerful signal about which techniques actually work — as skilled competitors develop state-of-the-art solutions, while overfitting is mitigated through the use of private test sets, and many of the results and winning approaches are published openly.