Few video games have rules as easy to explain as Sokoban. A worker stands inside a warehouse filled with boxes. Each box has to reach a designated storage position. The worker can walk around the floor and push a box from behind, but cannot pull it back. Push a box into the wrong corner and the puzzle may be lost. Put every box where it belongs and the stage is complete.
From those simple rules came one of the most enduring puzzle designs in video game history.
Sokoban began as a personal programming experiment by Japanese creator Hiroyuki Imabayashi during the early years of the home computer industry. It grew from a small BASIC program shared among friends into the foundation of Thinking Rabbit, spread across Japanese computers and consoles, crossed into Western markets during the late 1980s, and eventually became important enough that computer scientists used it to study artificial intelligence and computational complexity.
Its history is also a reminder of how easily the origins of early computer games can become blurred. Even the date of Sokoban’s first commercial release is disputed among surviving sources. What is much clearer is how the game itself emerged, and how an ordinary action, pushing an object across a screen, became the center of an extraordinary puzzle.
Before Sokoban
Hiroyuki Imabayashi did not begin his career with a long established ambition to become a professional game designer. His path into computers was far more accidental.
In a 1983 interview published in the Japanese computer magazine LOGiN, Imabayashi recalled that a friend brought a Sharp MZ computer to his home around 1980. The machine was heavy enough that the friend left it behind rather than immediately carrying it home. Imabayashi became fascinated. Although he knew little about computers at the time, he spent roughly three days and nights experimenting with the machine and its manual.
The first game he programmed was a simple action game involving a parachuting character trying to reach a landing position. He soon acquired a NEC PC-8001 and continued experimenting, producing several more games while still treating programming primarily as a hobby.
Computers were still young enough that an independent hobbyist could enter the field without an established development studio, a specialized degree, or an expensive production team. For Imabayashi, the machine was something to explore.
That experimentation eventually produced a puzzle unlike the other games he had made.
The Spring of 1981
Imabayashi’s own account places the birth of Sokoban in the spring of 1981.
Writing years later for the official Sokoban website, he remembered the season through the image of cherry blossoms. He explained that he had become interested in the movement of a character pushing luggage and decided to build a puzzle around that action using BASIC. The central idea was that the objects the player needed to put away would themselves interfere with the process of putting everything away.
The program came together relatively quickly. Designing good puzzles did not.
Imabayashi completed several stages and invited friends and acquaintances to his home to play them. Part of the entertainment for the designer came from watching them struggle with the problems he had created. At this stage there was no software company, no international publishing agreement, and apparently no expectation that Sokoban would become a commercial product. It was a hobby project meant to challenge people Imabayashi knew.
Later correspondence with Imabayashi provides even more detail. He recalled making the prototype in BASIC on his PC-8001, probably in spring 1981. The earliest version used extremely simple computer characters rather than the more recognizable graphics of later releases, and the prototype itself no longer survives.
The same correspondence also helps correct one story that has circulated about the game’s creation. Sokoban was not produced for a computer game contest. Imabayashi explicitly rejected that claim and said he had made the game simply as a hobby.
Aldebaran and the Idea of Pushing
The inspiration behind Sokoban’s most important mechanic can also be traced more directly than is possible for many games from the period.
When asked what gave him the idea, Imabayashi identified a Hudson computer game called Aldebaran. In that action game, luggage could be moved and used as a barrier against radiation. Imabayashi became interested in the act of moving the luggage itself and transformed that action into the basis of a puzzle.
Instead of pushing an object because it helped the player survive an action game, pushing became the entire problem.
That distinction was crucial. The player would not simply move boxes out of the way. Every push altered the structure of the puzzle. Because the character could push but not pull, some movements could not easily be reversed. A box placed against the wrong wall or pushed into the wrong corner might permanently destroy the solution.
The result was a game in which the player was constantly shaping the problem while attempting to solve it.
Aldebaran itself is documented in the January 1980 issue of Monthly Micom, providing contemporary evidence for the game Imabayashi later remembered as his inspiration. His explanation makes Sokoban an especially interesting example of game design evolving through reinterpretation. A supporting mechanic from one game became the central idea of another.
From a Record Store to Thinking Rabbit
The transformation of Sokoban from hobby project to commercial game happened through another unlikely chain of events.
Imabayashi’s wife’s family operated a record store. A computer section was eventually established there, giving customers access to machines including NEC computers. Imabayashi’s early version of Sokoban was among the software shown to people around the store.
Accounts differ slightly on exactly who first recognized its commercial potential. The 1983 LOGiN interview describes a customer seeing the game and commenting that it could become a hit. Imabayashi’s later official account remembers a salesman from a newly established software distribution company seeing Sokoban and urging him to turn it into a product.
Whichever encounter was decisive, the effect was the same. Imabayashi began preparing Sokoban for sale.
Using the PC-8801 available at the store, he improved the graphics, finalized the design, and expanded the commercial game to twenty stages. The 1983 interview presents an operation that would now be recognized as remarkably independent. Imabayashi duplicated the cassette tapes himself, created the packaging, and handled the business side rather than immediately turning the project over to an established publisher.
Surviving photographs from the period show the original cassette tapes and Imabayashi’s graph-paper level designs. Before sophisticated level editors became part of the series, the puzzles themselves were planned manually.
The success of that one game helped transform an amateur programmer into the founder of a software company.
Thinking Rabbit was born.
When Was Sokoban Actually Released?
The commercial release date presents one of the most interesting problems in reconstructing Sokoban’s history.
The official Falcon chronology lists Sokoban for the PC-8801, FM-7, and PC-8001mkII in December 1982. Imabayashi’s own retrospective account likewise states that he established Thinking Rabbit and released Sokoban in 1982.
A contemporary source complicates that chronology.
The December 1983 LOGiN profile introduces Thinking Rabbit by saying that Sokoban had been released in May. Since the article was published in December 1983 and described the company as only about six months old, the wording points toward a 1983 commercial timeline.
The Game Preservation Society creates another variation. Its preserved PC-8801 entry dates the Thinking Rabbit release to August 1983.
It is possible that these dates refer to different versions, stages of distribution, or later cataloging practices, but the surviving evidence does not provide a simple resolution. Rather than pretending the discrepancy does not exist, the safest historical conclusion is that the noncommercial game unquestionably existed by 1981, while Sokoban moved into commercial publication during the 1982 to 1983 period. Falcon and Imabayashi identify 1982 as the beginning of the commercial series, while important contemporary and preservation sources preserve evidence for 1983 dates.
Early video game history is often built from advertisements, magazines, surviving packages, memories, and corporate records rather than one perfect release ledger. Sokoban demonstrates exactly why those sources sometimes need to be presented together instead of reduced to a single date.
A Simple Game That Punished Careless Thinking
The brilliance of Sokoban was not technological complexity.
Imabayashi himself understood that the underlying program was comparatively simple. What mattered was the idea and the design of the individual stages. In his 1983 interview, he compared the process of creating Sokoban problems to composing tsume shogi, Japanese chess puzzles in which carefully arranged positions force the player to reason through a sequence of moves.
A contemporary December 1983 article in MICOMGAMES recognized the same quality. The magazine considered the concept easy to learn but emphasized how much thinking the puzzles required, comparing the intellectual demand to games such as Go and shogi. It also described the game as difficult to stop playing once someone began.
That depth came largely from one restriction: boxes could be pushed but not pulled.
In many puzzle games, a bad move costs time. In Sokoban, a bad move can change a solvable position into an unsolvable one.
Push a box into a corner that is not a goal and it may never move again. Place two boxes so that neither can be approached from the necessary direction and the stage can collapse. A movement that appears helpful in the short term may quietly eliminate the only path needed several minutes later.
The player therefore has to think not only about where a box should go, but about where the character must stand in order to push it there.
Every box is both cargo and obstacle.
From Twenty Levels to a Puzzle Community
Imabayashi originally designed the game’s stages himself, but Sokoban soon demonstrated that its basic rules could support far more puzzles than one designer could create alone.
In August 1983, PC Magazine published a special Sokoban edition containing ten additional problems supplied by Thinking Rabbit. Imabayashi later explained that designing stages involved considerable trial and error, leaving behind numerous puzzles that were not selected for the original commercial release.
The idea expanded further with Sokoban 2.
During his 1983 LOGiN interview, Imabayashi was already discussing the sequel. Unlike the first game, some of its maps were being created by another person. More importantly, he planned to include a utility allowing players to construct their own stages. To Imabayashi, having multiple people invent puzzles made sense because Sokoban functioned much like other composed logic problems.
The official series chronology places Sokoban 2 releases across several Japanese computer systems in 1984. The series was becoming more than a fixed package of twenty puzzles. It was becoming a system through which new puzzles could continually be invented.
That capacity for user-created levels would become one of the reasons the design survived long after its original hardware disappeared.
A Japanese Computer Game Bestseller
The early sales figures show that Sokoban was not merely admired by a small group of puzzle enthusiasts.
A June 1984 POPCOM feature reported approximately 22,000 copies sold. By the following month, another POPCOM article reported that the original game had exceeded 25,000 copies in roughly a year. The magazine placed that number in context by explaining that Japanese computer games commonly sold only a few thousand copies and that reaching 10,000 could qualify a title as a bestseller.
By March 1985, contemporary reporting placed the total above 30,000.
Those figures represented only part of what was becoming a much larger series.
ASCII brought Sokoban to the MSX standard during 1984, while other versions appeared across Japan’s fragmented home computer market. Falcon’s official chronology documents releases for systems including the Sharp X1, MZ series, PC-6001, PC-9801, MSX, Sega hardware, and later Nintendo’s Famicom Disk System.
The puzzle was unusually well suited to being moved between machines. Its identity did not depend on fast scrolling, elaborate animation, or hardware intensive action. As long as a system could display walls, boxes, targets, and a character moving around a grid, the essential design remained intact.
That portability helped Sokoban escape the lifespan of any single computer.
From Japanese Computers to Consoles
By the middle of the 1980s, Sokoban was moving well beyond its original cassette tape origins.
ASCII’s MSX versions appeared in 1984. Sega published a version for its SG-1000II, SC-3000, and Mark III family in December 1985. In July 1986, ASCII released Namida no Sokoban Special for Nintendo’s Famicom Disk System. Later versions appeared on the Game Boy, Mega Drive, PC Engine, Game Gear, Super Famicom, PlayStation, Windows, Macintosh, smartphones, and eventually modern downloadable platforms.
The graphic appearance changed repeatedly. So did publishers, characters, packaging, and the number of included stages.
The underlying puzzle survived.
That is an important distinction in the history of video games. Many early titles are closely tied to the hardware on which they first appeared. Sokoban became something closer to a design language. Its rules could move from computer to computer without losing their identity.
Soko-Ban Goes West
By the late 1980s, the game was ready for a much wider international audience.
Spectrum HoloByte brought the game to North America as Soko-Ban, with versions for IBM compatible computers, the Commodore 64, and Apple II appearing around 1987 and 1988. Falcon’s official chronology places the Spectrum HoloByte releases in 1988. Contemporary American material, however, shows the game being advertised and reviewed around the transition from 1987 into 1988, creating another small dating wrinkle in the game’s international history.
The Western edition reached reviewers who had little reason to care about how technically simple the original program had been.
What mattered was the puzzle.
In May 1988, Roy Wagner reviewed Soko-Ban alongside Tetris for Computer Gaming World. Other reviews appeared in Dragon, Compute!, ACE, The Games Machine, and European computer magazines. Contemporary reviewers repeatedly returned to the same qualities Japanese writers had noticed several years earlier: the rules were straightforward, the challenge was substantial, and the game could become remarkably difficult to put down.
By this stage, contemporary industry reporting said Sokoban had sold hundreds of thousands of copies in Japan before or around the time of its American expansion.
A small BASIC puzzle made for friends had become an international software property.
Deadlocks Become a Computer Science Problem
The same feature that frustrated players eventually attracted researchers.
Sokoban is filled with what computer scientists call deadlocks. A deadlock occurs when a sequence of apparently legal moves leaves the puzzle in a state from which completion has become impossible.
Humans often learn to recognize simple examples quickly. A box trapped in the wrong corner is obviously useless. More complicated deadlocks can be much harder to detect because several boxes, corridors, and required pushing positions interact with one another.
For an artificial intelligence system attempting to solve Sokoban, this creates an enormous problem. The computer can spend huge amounts of effort exploring sequences of moves that ultimately lead nowhere.
During the 1990s, researchers began using Sokoban as a serious test problem for artificial intelligence and search algorithms.
In 1997, Joseph Culberson showed that generalized Sokoban is PSPACE-complete, formally demonstrating that the apparent simplicity of pushing boxes conceals enormous computational complexity. Researchers Andreas Junghanns and Jonathan Schaeffer subsequently used the game to study single-agent search, including methods of detecting and avoiding deadlock positions.
Their work on the Rolling Stone solver turned the warehouse puzzle into a laboratory for studying how computers reason through large search spaces.
The strange result was that a game conceived by a hobby programmer in 1981 had become, less than two decades later, a problem difficult enough to interest theoretical computer scientists and artificial intelligence researchers.
Millions of Copies and Decades of Sokoban
Sokoban continued appearing long after the Japanese computer market that created it had been transformed.
Falcon’s official history records releases through PCs, consoles, mobile phones, smartphones, television data services, Nintendo Switch, PlayStation systems, and other platforms across several decades.
In 2018, Chukyo Television Broadcasting described Sokoban as a series with more than 4.1 million copies sold worldwide since its original release. That figure should be understood as a corporate claim rather than an independently audited lifetime total, but it gives some sense of how far the game had traveled from Imabayashi’s original BASIC experiment.
Its influence extends farther than official sales.
Thousands of unofficial box-pushing games have borrowed the basic structure. Puzzle creators continue to design new stages. Dedicated players study optimal solutions. Computer programs compete with increasingly difficult collections. The word Sokoban itself is now commonly used by players and developers to describe an entire style of pushing puzzle.
Few games achieve that kind of transformation. A title may be successful, influential, or frequently copied. Sokoban reached the point where its name could describe the kind of puzzle it invented.
Why Sokoban Matters
The history of Sokoban is not the story of spectacular hardware or a huge development studio.
It is the story of an idea.
Hiroyuki Imabayashi saw an action in another computer game, pushing luggage across a screen, and wondered what would happen if that movement became a puzzle. He programmed an experiment, invited friends to struggle through it, polished the concept when others recognized its potential, duplicated cassette tapes, designed levels on graph paper, and built a software company around the result.
The rules scarcely needed to change afterward.
Push the boxes.
Do not trap them.
Think about where you will need to stand next.
That was enough.
From a PC-8001 hobby program in 1981 to Japanese bestseller, international computer release, console series, artificial intelligence research problem, and worldwide puzzle format, Sokoban became one of the clearest examples in video game history of how simple mechanics can produce almost unlimited complexity.
Its warehouse may be small, but players have been getting lost inside it for more than four decades.
Sources & Further Reading
Falcon Co., Ltd. “倉庫番の歴歴史 [History of Sokoban].” Sokoban Official Site. Accessed October 13, 2026. https://www.sokoban.jp/history.html
Imabayashi, Hiroyuki. “Greeting.” Sokoban Official Site. Accessed October 13, 2026. https://www.sokoban.jp/greeting.html
Aguilera, Carlos Montiers. “My Conversation with Mr Hiroyuki Imabayashi.” Soko Boxes. Accessed October 13, 2026. https://sokoboxes.com/articles/my-conversation-with-mr-hiroyuki-imabayashi/
“Aldebaran.” Monthly Micom, January 1980, 22–28. Internet Archive. https://archive.org/details/micom-1980-01/page/22/mode/2up
“Thinking Rabbit: 1983 Developer Interview.” Originally published in LOGiN, December 1983. Translated by Shmuplations. Accessed October 13, 2026. https://shmuplations.com/thinkingrabbit/
“倉庫番 [Sokoban].” LOGiN, August 1983, 125. Internet Archive. https://archive.org/details/login-1983.08/page/124/mode/2up
“倉庫番[番外編] [Sokoban Extra Edition].” PC Magazine, August 1983, 52–53. Internet Archive. https://archive.org/details/pc-magazine-aug-1983-sokoban-extra-edition/page/n1/mode/2up
National Diet Library. “PC Magazine.” NDL Search. Accessed October 13, 2026. https://ndlsearch.ndl.go.jp/books/R100000002-I000000043411
“PART 2 脳みそを酷使するのである.” MICOMGAMES 1, no. 1 (December 1983): 38. Internet Archive. https://archive.org/details/micomgames-magazine-dec-1983-sokoban/page/3/mode/2up
“ソフトハウス訪問 [Visiting a Software Company].” POPCOM, June 1984, 131. Internet Archive. https://archive.org/details/popcom-198406/page/130/mode/2up
“倉庫番トラの巻 [Sokoban Strategy Guide].” POPCOM, July 1984, 101. Internet Archive. https://archive.org/details/popcom-198407/page/100/mode/2up
“作者が語る自信のニューソフト [The Creator Presents His New Software].” POPCOM, March 1985, 29. Internet Archive. https://archive.org/details/popcom-198503/page/28/mode/2up
“倉庫番2 [Sokoban 2].” LOGiN, July 1985, 76. Internet Archive. https://archive.org/details/logi-n-july-1985/LOGiN%20-%20July%201985/page/n77/mode/2up
“こうなるとアウトだよ [If It Gets to This, It’s Unsolvable].” Famicom Tsūshin, July 1986, 30. Internet Archive. https://archive.org/details/famitsu-0003/page/30/mode/2up
Tanaka, Junji, Hiroyuki Imabayashi, and M. Ishii. THE 倉庫番 [The Sokoban]. SCALE, 1986. ISBN 4-88239-606-8. https://sokoboxes.com/pdfs/sokoban_wikipedia.pdf
Spectrum HoloByte. Soko-Ban. Game manual. 1988. Internet Archive. https://archive.org/details/stx_Spectrum_HoloByte_Soko-Ban/page/n4/mode/1up
Tandy Corporation and Spectral Associates. Soko-Ban. Color Computer game manual. 1988. https://colorcomputerarchive.com/coco/Documents/Manuals/Games/Soko-Ban%20%28Tandy%29.pdf
“Review of Soko-Ban.” Computer Entertainer 6, no. 9 (December 1987). https://www.ataricompendium.com/archives/newsletters/video_game_update/computer_entertainer_dec87.pdf
“Soko-Ban.” Happy Computer, January 1988, 86. https://stcarchiv.de/happy-computer/1988
“Soko-Ban.” The Games Machine, no. 5 (April 1988): 66. Internet Archive. https://archive.org/details/thegamesmachine-magazine-05/page/n65/mode/2up
Lesser, Hartley, Patricia Lesser, and Kirk Lesser. “The Role of Computers.” Dragon, no. 132 (April 1988): 80–85. https://sokoban-jd.blogspot.com/2014/05/the-original-collection.html
Wagner, Roy. “Puzzling Encounters: Two Titles from Spectrum HoloByte’s International Series.” Computer Gaming World, no. 47 (May 1988): 42–43. https://mirrors.apple2.org.za/ftp.apple.asimov.net/documentation/magazines/computer_gaming_world/Computer%20Gaming%20World-1988_05_issue47.pdf
Cooke, Steve. “Soko-Ban.” ACE: Advanced Computer Entertainment, no. 8 (May 1988): 52. Internet Archive. https://archive.org/details/ace-magazine-08/page/n51/mode/2up
Wierda, Brian. “Soko-Ban.” Compute! 10, no. 6, issue 97 (June 1988): 57. Internet Archive. https://archive.org/details/1988-JUn-compute-magazine
Low, Lafe. “News Line: Made in Japan.” inCider, November 1988, 14–15. Internet Archive. https://archive.org/details/inCider_1988-11/page/n15/mode/1up
Ellison, Carol. “Special Report: Why Japan Can’t Write Software.” PC Computing 1, no. 5 (December 1988): 113–14. Internet Archive. https://archive.org/details/PC_Computing_1988_12/page/n116/mode/1up
Thinking Rabbit. Official website. Accessed October 13, 2026. https://thinkingrabbit.jp/
Chukyo Television Broadcasting. “『倉庫番』シリーズ36周年記念作品に関する発表.” Press release, July 10, 2018. https://www.atpress.ne.jp/news/160010
Game Preservation Society. “Sokoban.” Game catalog. Accessed October 13, 2026. https://www.gamepres.org/catalog/?id=8811007241004
Game Preservation Society. “Sokoban.” PC-8801 Game Library. Accessed October 13, 2026. https://www.gamepres.org/pc88/library/1983/soko.htm
Generation MSX. “Sokoban.” Accessed October 13, 2026. https://generation-msx.nl/software/thinking-rabbit/sokoban/release/203
Generation MSX. “Sokoban Tool Kit.” Accessed October 13, 2026. https://www.generation-msx.nl/software/ascii-corporation/sokoban-tool-kit/536/
Culberson, Joseph C. “Sokoban Is PSPACE-Complete.” Technical Report TR97-02, Department of Computing Science, University of Alberta, 1997. https://web.archive.org/web/20240528042859/http://cl-informatik.uibk.ac.at/teaching/ss07/alth/material/culberson97sokoban.pdf
Junghanns, Andreas, and Jonathan Schaeffer. “Sokoban: A Challenging Single-Agent Search Problem.” University of Alberta, 1997. https://webdocs.cs.ualberta.ca/~jonathan/publications/ai_publications/soko.pdf
Junghanns, Andreas, and Jonathan Schaeffer. “Single-Agent Search in the Presence of Deadlocks.” In Proceedings of the Fifteenth National Conference on Artificial Intelligence, 419–424. 1998. https://cdn.aaai.org/AAAI/1998/AAAI98-059.pdf
Dor, Dorit, and Uri Zwick. “SOKOBAN and Other Motion Planning Problems.” Computational Geometry 13, no. 4 (1999): 215–228. https://www.sciencedirect.com/science/article/pii/S0925772199000176
Junghanns, Andreas, and Jonathan Schaeffer. “Sokoban: Enhancing General Single-Agent Search Methods Using Domain Knowledge.” Artificial Intelligence 129, nos. 1–2 (2001): 219–251. https://www.sciencedirect.com/science/article/pii/S0004370201001096
Hearn, Robert A. Games, Puzzles, and Computation. PhD diss., Massachusetts Institute of Technology, 2006. https://citeseerx.ist.psu.edu/document?doi=4f62b15db2a5451837d4c7fffc6ff63fca972f49&repid=rep1&type=pdf
MobyGames. “Soko-Ban.” Accessed October 13, 2026. https://www.mobygames.com/game/1715/soko-ban/