Other programmers at MIT improved this version of the game, which was also playable between people at different universities over the nascent ARPANET.

There are reports that the Defense Advanced Research Projects Agency (DARPA) at one point banned the game from the ARPANET due to its popularity.

The maze layout is represented by a grid of spaces that are either empty or solid and form a flat plane containing walls of equal height.

Other players in the maze are displayed as the letters of their usernames along with an indicator of which direction they are looking; later versions of the game replaced this with the image of an eyeball.

Colley was developing a method of determining which vertices of a three-dimensional object would not be visible to a viewer and then not drawing them on the screen, thereby displaying a 3D model that looked solid rather than see-through.

[4] Thompson brought paper tapes of code for several programs from NASA Ames to MIT in February 1974, including Maze.

Thompson worked on the PDS-1 code that allowed for more players, the visuals for the bullets, the score-keeping, the ability to see a top-down view of the maze, and a cheat command to move through walls.

[7] As users had to reserve time on the terminals due to the limited availability, some players would go to the lab in the middle of the night in order to play the game.

According to Lebling, the first multiplayer game between institutions was between students at MIT and the University of California, Santa Cruz, although the slow speed of the network left the non-MIT players at a disadvantage.

The first was partially developed by Thompson himself; in the fall of 1976 he took an electrical engineering digital electronics design class, in which he had to do a group project with Mark Horowitz and George Woltman.

[1] In this version, the maze was a 16 by 16 by 16 cube with no gravity in which the player could move up and down just as they did forward and back, as they found it easier to create hardware that did not need to treat the floor and ceiling differently than other sides.

After Guyton moved to Xerox, the pair felt that the game would be suited to the Alto and could be improved on there, and Wahrman got copies of the PDP-10 and PDS-1 code.

They rewrote the game entirely in the Mesa programming language and were assisted by several other Xerox employees, including Steven Hayes, Bill Verplank, Jim Sandman, and Bruce Malasky.

Eventually, it migrated to MIT, Stanford, and Carnegie Mellon University, which were some of the few non-Xerox locations that owned Xerox Alto computers.

[13][14][15] It has additionally been credited with a variety of other firsts, such as level editing due to Lebling's editor, observer mode and radar from the top-down view, and avatars from the representation of other players.

[1] In 1987, MacroMind released a version of the game for the Apple Macintosh titled Maze Wars+, which was playable on the AppleTalk local network by up to 30 players.

[1] Several other games based on the Maze concept, with a variety of graphical styles and differences from the original versions, were released in the 1980s and 1990s.

These include MIDI Maze for the Atari ST by Xanth Software in 1987, which was ported as Faceball 2000 to the Game Boy, Super Nintendo Entertainment System, and Game Gear; Oracle Maze, a demo application at the Interop 92 conference to demonstrate Oracle's networking technology connecting many different companies' computers; MazeWars for NeXTSTEP by Mike Kienenberger in 1994, and MazeWars for Palm OS by IndiVideo in 1998.