Michael Buro

Michael Buro is a German computer scientist whose game-playing programs combine heuristic search with machine learning. His Othello program Logistello, the subject of his 1994 PhD at the University of Paderborn, defeated world champion Takeshi Murakami 6–0 in August 1997. After joining the University of Alberta in 2002, he worked on real-time strategy games and the card game Skat, built the Skat program Kermit with his students, and later paired search with convolutional networks and transformers. He retired in 2024 and is a professor emeritus of computing science at Alberta, where his research on planning with hidden information continues.

Now at: University of Alberta
Born: 1965
Works: 116
Citations: 2,573
h-index: 32
Badges: Bridge
Loyal

Years link to their sources. Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

1990–1994

Degrees at Aachen and Paderborn

A German computer scientist, he earns an M.Sc. in computing science at the Technical University of Aachen in 1990 with a diploma thesis on busy beavers. He completes a PhD at the University of Paderborn in 1994 with work on the Othello program Logistello.
1992–1998

Building Logistello

From 1992 to 1998, he develops Logistello, whose evaluation features are combined by regression and whose opening book learns from earlier games. He creates ProbCut, a selective extension of alpha-beta search published in 1995, and improves the program while working at the NEC Research Institute (NECI).
1997–1998

Logistello beats Takeshi Murakami

In August 1997, Logistello defeats Othello world champion Takeshi Murakami 6–0. Its evaluation function tunes over a million numerical parameters with linear regression, and the program retires from tournament play in January 1998.
2002

Joining the University of Alberta

He joins computing science at the University of Alberta in 2002. In November 2002, he releases Logistello's source code under the GPL so that others can study the implementation of a strong game-playing program.
2002–2012

ORTS and StarCraft AI competitions

In 2002, he describes ORTS, a hack-free real-time strategy game environment that hosts AI competitions from 2006 to 2009. In 2011, the University of Alberta's RTS Game AI Research Group organizes the second annual AIIDE StarCraft AI Competition, with Buro as its contact. With David Churchill, he publishes faster build-order search for StarCraft in 2011 and an AI Magazine survey of RTS game AI competitions in 2012.
2010–2019

Kermit and Skat AI

With his students, he builds Kermit, a computer program that plays the three-player card game Skat. In 2010, with Jeffrey Long, Nathan Sturtevant and Timothy Furtak, he studies when Perfect Information Monte Carlo search plays imperfect-information games well. In 2019, he coauthors two Skat methods, one sampling hidden cards with a deep neural network trained on human games and one inferring them through player modelling, which strengthens Kermit.
2017

Convolutional networks for RTS games

In 2017, with Nicolas Barriga and Marius Stanescu, he pairs a deep convolutional neural network that picks abstract actions with game-tree search for low-level tactics. In microRTS, the combined agent wins more often than either component alone and than other state-of-the-art agents.
Since 2024

Professor emeritus at Alberta

He retires from the University of Alberta faculty in 2024 and is a professor emeritus of computing science. His research on planning with hidden information continues with GO-MCTS, which trains transformers through self-play for Skat and Hearts, in April 2024, and with Belief Embedding Tree Search, coauthored with Michael Bowling and Nathan Sturtevant, in October 2026.