The one click bug on Windows 95
Early versions of Windows Minesweeper did not check if the first click had won the game. The timer would continue until the player flagged or clicked a mine to stop the clock.
The earliest known example was a Beginner game played by Damien Moore (Canada) on 26 March 2001. His second click hit a mine and when he inspected the board the first click had solved the game. He played another 3BV 1 board on 24 April 2002 and this time was forced to click a mine to stop the clock. This suggested that Minesweeper did not check whether your first click had won the game. However, other players reported completing 3BV 1 games with a single click including Roelof Smit (Netherlands) on 14 March 2002 and Perry Nock (Australia) on 7 July 2002.
Matt McGinley (USA) encountered a 3BV 1 Beginner game in September 2001 and eventually clicked on a mine after 5 seconds. He investigated and was able to prove that one mine had shifted to the top left corner turning a 3BV 3 game into a 3BV 1 game because Windows Minesweeper ensures the first click is safe. He reproduced the bug on a 16x16 board with 10 mines1.
Tim Kostka (USA) decided to check whether true 3BV 1 boards exist in Windows Minesweeper. He had previously demonstrated that the 8x8 Beginner grid produces fewer than 50,000 unique boards and now he proved there were no naturally occuring 3BV 1 games2. Instead, he found 5 boards which converted to 3BV 1 games when your first click was on a specific mine. However, only 4 of these boards can be completed in one click making the odds of having the one click bug about 1 in 800,000.
Damien scanned all 48,624 unique 8x8 Beginner boards in 2020 and checked games submitted to the rankings. This confirmed the one click games in the two Beginner cycles were boards #81 (Damien) and #7120 (Tim) in the even cycle and boards #7694 (Elmar) and #24019 (Matt) in the odd cycle. The second board is interesting because the true board has a mine in both the top left corner (0,0) and adjacent cell (1,0) so the shifted mine ends up in the third column (2,0).
In Windows XP the 9x9 Beginner grid does not have board cycles resulting in an unknown number of true 3BV 1 games. The timer always stops in Windows XP. This suggests the one click bug is the result of the mine shift combined with a single click. For example, in Matt's article he reproduced the bug on Intermediate with 10 mines and his picture shows a shifted mine in the top left corner. In contrast, the majority of 3BV 1 games on this Intermediate custom level do not mine shift so the timer stops.
But that is not the full story! What about a board played by Yoni Roll (Israel) on 12 August 2002 or the 8x8 Beginner games where the clock stopped? These games are not in either Beginner board cycle but this is possible becase it usually takes several games before you enter a board cycle3. However, this still does not explain the games where the timer stopped.
Source code for Windows NT 4.0 (from 1996) was leaked in 2004 including the source files for Windows Minesweeper. On your first click the StepSquare function checks if the cell contains a mine then moves the mine to the first available empty cell starting from the top left corner. The function does not check if your first click won the game! This means it is impossible to win a one click shifted mine game and have the timer stop. Consequently, the 8x8 Beginner games where the timer did stop were made with two clicks (with the first click shifting a mine but being a number).
The one click bug was eventually fixed with the release of Windows Vista in 2006.
Pictures
References
| 1 | 1 Second Theory - McGinley suggests that the one click bug is caused by first click mine shifts. |
| 2 | One Click Bug - Kostka confirms there are no 3BV 1 boards in the two Beginner board cycles. |
| 3 | Board Cycles - Kostka proves that it can take several games to "fall into" a board cycle. |
Article created 28 October 2020 by Damien Moore and updated 5 August 2021.