{
  "id": "lee-sedol",
  "title": "Lee Sedol",
  "lang": "en",
  "source_lang": "en",
  "summary": "Lee Sedol is a South Korean Go professional whose major tournament victories and 2016 match against AlphaGo made him an internationally prominent figure.",
  "category": "society",
  "subcategory": "culture",
  "type": "person",
  "aliases": [
    "Lee Se-dol",
    "Yi Se-dol",
    "I Se-dol"
  ],
  "generated_at": "2026-11-02",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/lee-sedol/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/lee-sedol/",
      "json": "https://aiwiki.page/zh/lee-sedol.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Lee Sedol (born 2 March 1983, Bigeum Island, South Jeolla Province, South Korea) is a South Korean professional Go player. A professional from 1995 until his retirement in 2019, he reached the highest professional rank, ninth dan, and won 18 major international individual championships. His tactical imagination, forceful fighting, and ability to turn apparently unfavorable positions into complicated contests made him one of the leading players of his generation.\n\nOutside the Go world, Lee became widely known through his five-game match against AlphaGo, the computer program developed by DeepMind, in Seoul in March 2016. He lost the match 4–1 but won the fourth game, producing a celebrated tactical move that exposed a weakness in the program’s evaluation. The encounter became a landmark in the public understanding of artificial intelligence, while also drawing attention to a human career already distinguished by more than a decade of elite competition."
    },
    {
      "heading": "Childhood and professional formation",
      "text": "Lee grew up on Bigeum Island in Sinan County, off South Korea’s southwestern coast. Go, known in Korean as baduk, was part of his family environment. He learned the game from his father, and his older brother, Lee Sanghun, also became a professional player. His early development thus began outside the major urban institutions that organized much of Korean professional Go, before advanced training brought him into the national competitive system.\n\nHe entered professional Go in 1995 at the age of twelve. In the Korean system, professional status is obtained through a highly competitive qualification process rather than simply by accumulating amateur tournament victories. Becoming a professional opens access to official competitions, but does not itself establish a player among the elite. Lee’s subsequent rise depended on results against experienced opponents in domestic leagues, title matches, and international qualifying events.\n\nHis development coincided with a period of exceptional Korean strength in international Go. Cho Hunhyun had helped establish South Korea as a major international force, while Lee Changho became the dominant championship player of the 1990s. For an emerging Korean professional, the domestic field therefore included players who were also among the strongest in the world. Success within Korea was not merely preparation for international competition; it frequently required defeating world champions.\n\nLee Sedol’s early reputation rested on aggressive, unconventional play. He was particularly dangerous in positions involving unsettled groups, complicated exchanges, and long tactical sequences. In Go, a group can remain strategically vulnerable even when it is not under immediate threat of capture. Lee showed an unusual ability to exploit such vulnerabilities, linking attacks in different parts of the board and finding resources that were difficult to assess through general principles alone.\n\nThe contrast between Lee Sedol and Lee Changho became a recurring theme in coverage of Korean Go. Lee Changho was associated especially with positional control, restraint, and precise endgame play, whereas the younger Lee was identified with tactical creativity and disruptive fighting. These descriptions were useful shorthand rather than complete accounts of either player. Lee Sedol’s eventual championship record required accurate judgment and disciplined conversion of advantages as well as spectacular attacks.\n\nHis rise also reflected the increasingly international structure of the profession. Elite players could no longer be assessed only through domestic titles. Regular events brought Korean, Chinese, and Japanese professionals into direct competition, providing repeated tests of a young player’s strength across different tournament formats and styles."
    },
    {
      "heading": "International breakthrough and championship record",
      "text": "Lee’s first major international individual championship came in 2002, when he won the Fujitsu Cup, defeating Yu Bin in the final. The tournament was one of the most important international events of its period. Winning it established Lee as more than a promising attacking player: he had demonstrated that his approach could succeed through a complete championship field.\n\nIn 2003 he won the LG Cup, defeating Lee Changho in the final, and successfully defended his Fujitsu Cup title. These results marked a decisive change in his standing. Victory over Lee Changho carried particular significance because the older player had been the central figure in international Go for much of the preceding decade. Lee Sedol was promoted to ninth dan in 2003, reaching the highest professional rank while still only twenty years old.\n\nProfessional dan ranks require some explanation. They are not a continuously updated world ranking, and the difference between an eighth-dan and a ninth-dan professional does not necessarily indicate a predictable difference in playing strength. Promotion rules have also changed over time and differ between national organizations. Lee’s ninth-dan status recorded his professional achievement; his repeated victories against elite opponents provided the stronger evidence of his competitive position.\n\nOver his career, Lee accumulated 18 major international individual titles. These included victories in the Fujitsu Cup, LG Cup, Samsung Cup, Toyota–Denso Cup, Chunlan Cup, and BC Card Cup. Such championships differed in field size, qualification rules, time controls, and final-match format. Winning across several of them demonstrated adaptability rather than mastery of a single competition’s conditions.\n\nHis three successive BC Card Cup victories, from 2008 through 2010, were a notable example of sustained success. He also continued winning major titles into the following decade, including the 2012 Samsung Cup. His championship career therefore spanned the transition from a period of pronounced Korean dominance to one in which Chinese professionals increasingly won the leading international events.\n\nThe total of 18 major titles refers to a conventional category used in professional Go records. It does not include every domestic title, team victory, invitational event, or exhibition match. This distinction matters because Go has no single championship whose winner automatically becomes an undisputed world champion. Several major international tournaments coexist, and assessments of the leading player depend on results across them. Lee’s standing rested on both the number of his victories and the length of time over which he remained a serious championship contender."
    },
    {
      "heading": "Rivalries and the changing competitive field",
      "text": "Lee’s career overlapped with several distinct generations of international Go. Early on, Lee Changho remained a central opponent and benchmark. Later, the Chinese player Gu Li became his most prominent generational rival. In the final years of Lee Sedol’s career, younger players including Ke Jie increasingly occupied the leading positions in international competition. His record consequently traces a broader redistribution of strength within the professional game.\n\nThe rivalry with Gu Li was especially notable because both players were associated with powerful fighting and deep tactical calculation. Their games often involved extensive exchanges in which territory, influence, and the safety of several groups had to be assessed together. Rather than reducing the contest to a single local battle, both were capable of sacrificing stones in one area to obtain an advantage elsewhere. Their encounters attracted interest not only for the results but also for the complexity of the positions they produced.\n\nIn 2014 they played a ten-game match sponsored by the Chinese company MLily. The format recalled the prestige of historical long matches in East Asian Go, in which repeated games allowed an extended comparison between leading players. Unlike a knockout tournament, a ten-game series gives each participant opportunities to recover from an individual defeat, adjust preparation, and respond to an opponent’s developing choices.\n\nLee won the series 6–2, bringing it to an end after eight games because Gu could no longer draw level. The result was an important late-career achievement against an opponent of comparable stature. It did not erase the balance of their wider rivalry or establish superiority in every context, but it provided a clear outcome in a deliberately extended head-to-head contest.\n\nBy contrast, Lee’s five-game final against Ke Jie in the second MLily Cup, completed in January 2016, ended in a 3–2 defeat. The match took place shortly before his encounter with AlphaGo. It showed that he remained capable of reaching the final of a major international championship, even as a younger generation was becoming increasingly prominent.\n\nThese rivalries also illustrate why national labels alone cannot explain professional Go. Korean and Chinese institutions supplied many of the strongest players, but individual careers depended on distinctive training histories, styles, and responses to particular opponents. Lee’s importance lay partly in his continuity across these changes: he competed against the established champions of his youth, became a defining rival for his contemporaries, and remained competitive with players whose careers began much later."
    },
    {
      "heading": "Playing style and technical strengths",
      "text": "Lee was especially associated with fighting positions in which straightforward territorial accounting was inadequate. On a standard 19-by-19 Go board, an apparently profitable local move may leave a group vulnerable or concede influence that matters much later. Lee excelled at identifying such delayed consequences. He could allow an opponent to gain something visible while creating tactical weaknesses that became decisive only after further exchanges.\n\nA central component of this style was reading: calculating possible sequences while accounting for liberties, captures, connections, and the survival of groups. Go differs from chess in that individual stones do not have different movement rules or fixed material values. A stone’s importance depends on its relationships to nearby and distant stones. Lee’s strongest combinations often exploited this contextual value, treating a group that appeared important as expendable while preserving less conspicuous stones with greater strategic potential.\n\nHis play frequently made use of aji, the latent possibilities remaining in a position. A stone or cutting point that has no immediate effect may acquire significance after the surrounding position changes. Removing such possibilities can simplify a game, but may also cost a move. Lee was adept at leaving unresolved features on the board and then using them when an opponent’s later commitments made them effective.\n\nHe was also renowned for creating resistance in inferior positions. A player who is behind may need to increase uncertainty rather than continue with locally efficient but predictable moves. Lee could complicate the game through sacrifices, forcing sequences, or attacks that obliged an opponent to make several accurate decisions in succession. This was not equivalent to playing random risks. The complications typically depended on concrete reading and on weaknesses already present in the position.\n\nDescriptions of Lee as an intuitive or creative player should therefore not obscure the technical basis of his success. Elite tactical play requires positional judgment: recognizing which fights are worth starting, how much territory can be conceded, and whether an attack will produce a lasting benefit. His championship victories also required endgame accuracy, time management, and the ability to win positions in which no dramatic combination was available.\n\nAt the same time, his style was not a universal solution to every opponent or position. Complex fighting gives opportunities to both players, and a mistaken assessment can turn an ambitious attack into an overextension. The enduring interest of Lee’s games comes from this interaction between imagination and verification. They offer examples of how an elite player constructs problems for an opponent while continually testing whether the resulting complications are favorable."
    },
    {
      "heading": "AlphaGo and the preparation for the Seoul match",
      "text": "Before AlphaGo, computer programs were much weaker at full-board Go than the strongest human professionals. The difficulty was not simply that the game had many possible positions. Evaluating a position was also challenging: territorial frameworks might never become secure territory, groups could survive through indirect threats, and the value of a move often depended on developments across the board.\n\nEarlier advances had come from Monte Carlo tree search, which estimates the promise of moves through repeated exploration of possible continuations. AlphaGo combined search with artificial neural networks. A policy network helped identify promising moves, while a value network estimated the likely outcome of positions. The system thereby reduced the number of variations requiring attention and improved its assessment of the positions reached.\n\nIts development used both professional game records and reinforcement learning through self-play. These processes served different functions. Human games supplied examples of strong play, while self-play allowed the system to improve by competing against versions of itself. The resulting program was not simply retrieving complete games from a database; it was selecting moves through learned evaluations and search.\n\nIn October 2015, AlphaGo defeated the European champion Fan Hui 5–0 in an even, full-board match. The result was announced publicly in January 2016 alongside a paper in Nature. Fan was a professional player, making the victory a major technical milestone. However, the strength difference between a European champion and one of the world’s most accomplished tournament players left an important question about the program’s ability at the highest level.\n\nThe match against Lee was arranged for 9–15 March 2016 at the Four Seasons Hotel in Seoul. It consisted of five games, with colors alternating and no handicap. Each player received two hours of main thinking time followed by three one-minute overtime periods. Chinese rules were used, with White receiving 7.5 points of compensation, or komi, for playing second. The advertised prize for winning the match was US$1 million.\n\nThe program that faced Lee was stronger than the version that had played Fan Hui. Lee nevertheless entered the encounter with confidence that human championship-level judgment would remain competitive. This expectation was understandable in the context of earlier computer Go, but it was tested almost immediately. The match became an unusually public evaluation of a fast-developing machine-learning system against a human expert whose qualifications were already established independently of the experiment."
    },
    {
      "heading": "The five games of March 2016",
      "text": "AlphaGo won the first game on 9 March, taking White against Lee’s Black. Lee resigned after 186 moves. The result immediately changed the interpretation of the match: the program was not merely capable of occasionally troubling a top player, but could sustain strong play through an entire serious game. Its choices also challenged commentators who initially interpreted some unconventional moves through familiar human expectations.\n\nThe second game, on 10 March, became famous for AlphaGo’s move 37, played as Black. It was an unusual fifth-line shoulder hit that many observers initially regarded as questionable. Subsequent play revealed its strategic effectiveness. The move became a widely cited example of the program selecting a strong continuation that was unlikely under conventional professional habits. AlphaGo won after 211 moves, taking a 2–0 lead.\n\nLee lost the third game on 12 March, which settled the best-of-five match in AlphaGo’s favor. The remaining games nevertheless continued. They were valuable as competitive encounters and as further tests of the program, since a match victory did not imply that every kind of position had been mastered. Lee’s response to the first three defeats was therefore examined closely by both Go professionals and the broader audience.\n\nIn the fourth game, on 13 March, Lee played White and produced the move most closely associated with the contest: White 78, a wedge inserted into Black’s central position. The move created difficult tactical problems and induced a sequence of errors by AlphaGo. Although the program did not immediately recognize the full deterioration of its position, it eventually resigned after White’s 180th move.\n\nWhite 78 was widely called a “divine move,” an honorific expression in Go commentary rather than a formal classification. Its importance lay in the concrete problems it created. The game showed that the program’s evaluations could fail in a complicated position and that its subsequent search did not reliably repair the mistake. Lee’s victory was therefore both a competitive achievement and a revealing case of a strong system performing poorly outside the lines it handled most successfully.\n\nFor the fifth game, played on 15 March, Lee took Black at his request. AlphaGo recovered from an early difficulty and won after 280 moves, completing a 4–1 victory. The final score established the program’s superiority in that particular match while preserving an important distinction between match strength and infallibility. Lee’s fourth-game win remained the only defeat in AlphaGo’s publicly documented matches against human professionals on an even board."
    },
    {
      "heading": "Late career and retirement",
      "text": "The AlphaGo match did not immediately end Lee’s professional career. He continued to participate in human competition, including domestic events and international tournaments. This continuation matters because the public narrative sometimes compresses his career into a single confrontation with a computer. In professional Go, however, he remained a competing player after March 2016 rather than becoming solely a symbolic representative of human intelligence.\n\nThe competitive environment was changing rapidly. Stronger Go programs became available for analysis, and younger professionals were already winning a growing share of major titles. Lee’s final years therefore combined two developments: the ordinary generational pressures of elite sport and a technological transformation in how the game could be studied. These were related in practice, but neither can be reduced entirely to the other.\n\nIn November 2019 Lee submitted his retirement notice to the Korea Baduk Association, ending a professional career that had begun in 1995. In public comments surrounding his retirement, he discussed the effect of increasingly powerful Go AI on his understanding of competitive achievement. His remarks concerned the difficulty of maintaining the former ideal of being the strongest when computer systems had moved beyond the human field.\n\nThose statements attracted attention because Lee had personally demonstrated both sides of the transition: he had lost a landmark match to AlphaGo, but had also found a way to defeat it once. They should not be treated as proof that AI made human Go meaningless, or as a complete explanation of every consideration behind retirement. His professional career had already lasted approximately twenty-four years, and his championship achievements substantially predated the computer match.\n\nIn December 2019 he played a three-game farewell series against HanDol, a Go program developed by the South Korean company NHN. He won the first game with a two-stone handicap and lost the following two. The handicap is essential to interpreting the result: an initial advantage given to one side changes the competitive conditions, so this series was not directly comparable to the even-board AlphaGo match.\n\nThe farewell event provided another public encounter between an experienced human player and a strong program, but it did not overturn the development demonstrated since 2016. By then, leading Go software had become an analytical resource for professionals rather than simply an opponent awaiting a decisive human challenge. Lee’s retirement thus occurred after computer strength had moved from a surprising result to a routine feature of the game’s professional environment."
    },
    {
      "heading": "Scientific and cultural significance",
      "text": "The Seoul match became a major public demonstration of advances in AI because its result could be understood without specialist knowledge of the underlying technology. Go had long been regarded as a demanding test of strategic judgment. Defeating a player with Lee’s record made the program’s strength visible in a way that benchmark scores or technical descriptions alone could not.\n\nThe comparison with Garry Kasparov and his 1997 defeat by IBM’s Deep Blue was frequent, but the systems and games differed. AlphaGo’s combination of learned policies, learned position evaluation, and search represented a different technical approach from the methods most strongly associated with Deep Blue. The two encounters belong to the same history of computer game-playing without being interchangeable examples of a single mechanism.\n\nThe match also affected Go practice. Strong AI systems encouraged professionals to reconsider opening patterns and positional judgments that had been supported by longstanding experience. Later programs, including AlphaGo Zero and other analysis engines, extended this change. The shift was not that every unconventional computer move was automatically correct, but that reliable analysis could reveal effective ideas outside established human preferences.\n\nLee’s fourth-game victory supplied an important counterpoint to simplistic descriptions of the technology. A system could outperform a leading professional overall and still make serious mistakes in a particular sequence. Its failure did not negate its general strength, just as its match victory did not demonstrate unrestricted reasoning ability in every domain. AlphaGo was developed for Go; the Seoul result was evidence about that task, not proof of a general solution to intelligence.\n\nThe encounter reached audiences through broadcasts, journalism, and the 2017 documentary *AlphaGo*, directed by Greg Kohs. The film followed the program’s development and the human participants around the match, helping preserve the tension between a scientific project and a professional contest. Lee’s expressions, preparation, defeats, and eventual victory made an abstract technological development legible as a sequence of human events.\n\nWithin Go, his historical standing remains grounded in a larger record: 18 major international titles, repeated success against leading contemporaries, and games valued for their tactical originality. The AlphaGo match added a distinct place in the history of computing. Its surviving game records allow both dimensions to be studied directly—Lee as a championship professional making difficult decisions over the board, and AlphaGo as a system whose striking strengths and specific failures became visible through those decisions."
    }
  ],
  "links": [
    "go-game",
    "alphago",
    "deepmind",
    "artificial-intelligence",
    "lee-changho",
    "ke-jie",
    "chess",
    "monte-carlo-tree-search",
    "artificial-neural-network",
    "reinforcement-learning",
    "machine-learning",
    "garry-kasparov",
    "deep-blue",
    "alphago-film"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}