The Atari Video Computer System was built to play Combat. That's it. Every piece of its hardware -- each chip, the iconic joystick, and the half-dozen toggles on the console -- was designed to run Combat perfectly. And Combat was an impressive game for 1977. It featured 27 different game types, each with single- and two-player options, and even with the option to handicap the better player. The story of the Atari VCS (later re-named the 2600), as told in Nick Montfort and Ian Bogost's Racing the Beam, begins there.
The book, part of a proposed series of "platform studies," tells the tale of the VCS by examining several important games throughout its life cycle. As the authors tell it, the console's defining hardware characteristic was the Television Interface Adaptor, which displayed the software's visual information at the same pace as the electron gun that draws the picture on a standard CRT television. It was a solution that made perfect sense for Combat, whose playing field was exactly one screen in size, and whose only moving elements were two vehicles and one or two projectiles.
But the more interesting story of the VCS is all the things it wasn't designed to do, and how programmers made it do those things anyway. There were no hardware shortcuts to create the multi-screen dungeon of Warren Robinett's Adventure. (In fact, one of my favorite bits of trivia in Racing the Beam is how the instruction manual for Adventure had to explain to players how to navigate between screens. It was that novel a concept.) The colorful neutral zone of Yars' Revenge was an innovative bit of programming trickery without precedent on the system. And until the advent of Activision, the first big third-party publisher, it was usually a lone programmer hashing out these games in a period of months or even weeks.
One of the most oft-cited examples of VCS failure is Tod Frye's adaptation of Pac Man. As Racing the Beam tells it, it's amazing that the game turned out as well as it did. In contrast to Combat, nothing about the VCS's design was suited to Pac Man. The intricate, asymmetrical pathways of the arcade version were too cumbersome for the system memory. The system wasn't supposed to be able to compute the movements of so many sprites at once. Most troubling for Pac Man, the TIA didn't even allow the display of circular objects. Frye's solutions to the challenges that faced him (in a timespan of only six weeks), frankly, seem like ingenuity at its finest.
Racing the Beam can be a bit technical in parts, but it requires no expertise to enjoy the stories of programmers finding new ways to squeeze more performance out of increasingly antiquated hardware. The jump from Combat to Yars' Revenge to Pitfall is so pronounced that it's almost like moving onto a different platform altogether. Turns out that programming isn't so different from playing games. First you learn the rules, and then you learn how to break them.