In a product development office in Hawthorne, California, some time in 1976, a Mattel engineer named George J. Klose watched a single red dot crawl up a calculator display and decided it was a car. That dot became Auto Race, the first handheld electronic game to run entirely on solid-state parts, with nothing mechanical in it beyond the switches and the on/off slider. It sold for $24.99, ran off a nine-volt battery, and the whole program fitted into roughly half a kilobyte of ROM.
- Maker: Mattel Electronics, Hawthorne, California. Model 9879, printed in Hong Kong
- Copyright date on the manual and case: 1976. First shown publicly at Toy Fair 1977
- Chip: Rockwell B6000, redesigned from the B5000 calculator part
- Program: 511 or 512 bytes of ROM, depending on which source you trust
- Display: red LEDs on a three-lane track, plus a two-digit lap-time readout
- Launch price: $24.99 (US)
None of that is the interesting part. The interesting part is that two engineers at two different companies had to talk a calculator into forgetting what it was for.
The engineer who taught a calculator to lie
Klose was a product development engineer at Mattel, and his idea began with something the rest of the building had been staring straight past for years. A calculator display is a grid of individually addressable segments. Light one and you have a dot. Light the next one along, kill the first, and you have a dot that moves. Do it fast enough and the person holding the thing will swear blind they are watching something travel.
He wanted the blips to be cars. The coin-op racers of the early 1970s were the reference point, the ones where a track came scrolling towards you and the whole skill was picking gaps, and Klose could see no good reason that idea needed a cabinet the size of a wardrobe to work.
Persuading anyone was the hard bit. Mattel was a toy company, and toy companies in 1976 did not commission custom silicon on the strength of an engineer’s hunch. So Klose built the argument instead of making it: a working demonstration of a blip moving across an LED display with no microprocessor anywhere in the circuit, just calculator hardware doing something nobody had designed it to do. That got him permission to carry on. It is a small thing and it is the entire reason the category exists.
Why Mattel had to ring Rockwell
Mattel could mould a case and print a box. What Mattel could not do was design an LSI chip, or shrink a circuit board far enough to disappear into a pocket. Klose and his manager, Richard Cheng, went looking for someone who could, and landed on the Microelectronics Division of Rockwell International, at that point one of the strongest names in handheld calculator chip design.
What the two of them carried into that meeting was, by Mark Lesser’s account, a single sheet of paper. ‘Mattel approached Rockwell with the idea of converting a calculator into a handheld game, and George Klose and Richard Cheng came up with a one sheet description of an auto race game,’ Lesser told Digit Press years later. One page. That was the specification for an entire product category, and nobody in the room yet knew whether it could be built at all.
Those documents survive. The Brian Sutton-Smith Library and Archives of Play at The Strong holds the game specification notes for Auto Race and Football, credited to George Klose and Mark Lesser and dated 1976.
Mark Lesser had never written a program in his life
Rockwell handed the job to Mark Lesser, who had come out of MIT with an electrical engineering degree and joined the Microelectronics Division in 1972 to design calculator chips. He was a circuit designer. He had never written a line of software.
‘It was up to me to do the actual chip redesign and to write the program from scratch,’ he said. ‘The design work included a multiplexing scheme for the display outputs.’
So he did both jobs at once. He took the B5000, an ordinary LED calculator part, and added a display driver multiplexing scheme so the segments could be scanned as a playfield rather than a row of digits, plus a custom sound driver for a piezo-ceramic speaker. Rockwell designated the result the B6000.
The sound arrangement is the part that still makes me wince on his behalf. There was no sound hardware in any meaningful sense: one output line, one piezo element and that was the lot. ‘The sound was implemented using a single output and a piezo-ceramic speaker,’ Lesser said, ‘one output line to generate sounds and music.’ Every beep in Auto Race is the program toggling that pin inside a timing loop written into the game code itself, which means the engine note and the crash tone were being paid for out of the same instruction budget as the collision detection. Play a sound and the game stops thinking for exactly as long as the sound lasts. Sit with that for a second, then consider how tight the budget was.
Lesser’s own figure is 511 bytes. ‘All the game logic, score keeping, timing and display was limited to 511 bytes of code,’ he said, and the exclamation mark in the original interview is his, not mine. Most secondary accounts round it up to 512, which is the ROM size of the B6000 part rather than the space he actually had for a program.
For scale, Riddle’s teardown work on the Mattel chips puts the later B61xx parts, the ones behind Football and Baseball, at 896 bytes. Lesser wrote the first game in the family with a little over half of that, in assembly language, having never programmed before, on a chip he was redesigning underneath himself as he went. The chips came back from the fab in unusual 42-pin quad in-line packages carrying almost no markings at all: a Rockwell logo, a date code, an ID. Nothing to suggest the thing inside had a job.
Wikipedia’s account has him spending eighteen months getting the code to fit.
He got there. His own site records one detail that says more about the work than any release date could: Auto Race was the only game program he ever wrote with no known bugs. He went on to Parker Brothers around 1980, wrote Frogger II: Threeedeep! for the Atari 2600 in 1984, and spent the 1990s on sports titles including the NHL series. None of them managed it.
What 511 bytes actually buys you
Pick the thing up and the constraint is visible before you switch it on. The playfield is a narrow black slot occupying maybe a third of a grey plastic case, with the three lane dividers silkscreened onto the plastic as white lines because painted lines cost nothing and LEDs cost money. The gear switch runs down the left in a colour ramp from yellow at 4 to red at 1. There is a chequered flag printed above the START and RESET slider. The steering control is a small ridged knob at the bottom right, and it moves your car between three positions, which is the entire steering model.
The rules, from the original Mattel instruction sheet: complete four laps in 99 seconds or less. Your car is ‘the blip of light that starts at the bottom of the track and moves UP to the top’, and the traffic starts at the top and comes down. The game will not start unless the gear switch is set to 1, a restriction the manual flags in bold because it is not a design flourish, it is what the code needed to initialise properly. Collide, and, in the manual’s careful phrasing, ‘you may be penalized back to the starting line losing valuable time’. The advice that follows is the best sentence in the document: ‘If you have a collision, switch driving lanes quickly for fewer penalties.’
Then there is the scoring table, which grades your four laps and does not soften the blow:
- Under 30 seconds: World Champion Driver
- 30 to 45 seconds: Professional Driver
- 45 to 55 seconds: Showing Potential
- 55 to 65 seconds: Still an Amateur
- 65 to 75 seconds: Stick To The Highways
- 75 seconds or more: Leave Car In Garage
Nobody writes toy copy like that any more, and the toy industry is poorer for it.
One more detail worth knowing if you own one. Behind the triangle-head screws on the back there is a potentiometer that sets the game’s clock speed, and the Handheld Museum records that it can roughly double or halve the pace. It was a factory calibration control, set to the middle and never mentioned to the customer. Adjustable difficulty existed in 1976. It just was not for you.
Was Auto Race really the first handheld electronic game?
Not quite, and the precise version of the claim is more impressive than the loose one.
Waco’s Electro Tic-Tac-Toe arrived in 1972 and is commonly cited as the first commercially available handheld electronic game. Its status is disputed for a good reason: it leaned on the players for the logic and the scoring. No electronic opponent, no scoring, not so much as a beep. It lit up when you told it to.
Auto Race’s claim is narrower and it holds. It was the first handheld game in which the electronics ran the whole thing: the opposition, the collision detection, the timer, the scoring, the sound. Everything that made it a game happened inside the chip. That is a genuinely different object from a box of switches and lamps, and it is the line the whole industry was built on the far side of.
Was Auto Race released in 1976 or 1977?
Both dates turn up in reputable places, and both are defensible.
The copyright line on the instruction sheet reads ‘Mattel, Inc. 1976. Hawthorne, California 90250’, and Wikipedia along with most secondary accounts gives 1976 as the release year. The Handheld Museum records it as designed in 1976 but not shown to the public until Toy Fair 1977, with retail following later that year, and notes that the copyrights simply reflect when the work was finished. The Strong holds a B6000 ROM code printout dated 13 February 1977, which is comfortably after the copyright date and comfortably before Christmas.
The tidy answer: 1976 is when it existed, 1977 is when you could buy one.
Football took the credit, and Mattel nearly talked itself out of both
Auto Race did well enough to justify a follow-up, and here the record genuinely disagrees with itself. Wikipedia has its sales exceeding expectations and convincing Mattel to fund Football. The Strong’s own museum blog has the opposite, describing Auto Race’s initial sales as underwhelming and crediting Football with the breakthrough. Both accounts come from serious sources and they cannot both be right.
What is not in dispute is what happened next. Football shipped in 1977 on the roomier B61xx chip through Sears, and it went off like a firework. By 1978 Mattel had spun up a dedicated Mattel Electronics Division to handle a business that had started as one engineer’s side project with a calculator chip. Mattel’s Michael Katz later described what these games had become as a ‘$400 million category’.
The pattern in all of it, though, is the one worth naming. The engineering worked on the first try and stayed working. The forecasting was wrong in both directions and repeatedly. Klose had to build a prototype to get taken seriously inside his own company, Lesser wrote a bug-free game on 511 bytes without ever having programmed, and the part of the business that kept getting it wrong was the part with the spreadsheets.
What the dot started
Mattel Electronics ran the same trick over and over for the rest of the decade: Missile Attack, Football, Baseball, Soccer, Hockey, a Battlestar Galactica Space Alert tie-in in 1978, Ski Slalom in 1980. Lesser built seven handheld games in his four years at Rockwell before moving on.
The idea travelled further than that. In 1983 the Soviet Union produced its own version as the Elektronika IER-01, which is a strange kind of compliment: a Hawthorne toy engineer’s calculator hack, reverse-engineered and reissued behind the Iron Curtain seven years later. Nintendo had already shipped the first Game & Watch in 1980, and the direct line from a red blip on a segment display to the Game Boy in 1989 is short enough to draw in one stroke.
Which leaves the object itself, and the thing I keep coming back to. There is a box in an archive in Rochester, New York containing a printout dated 13 February 1977: 511 bytes of hand-written assembly language for a machine whose only means of expression was a red dot in one of three positions. Out of that came the timer, the traffic, the four gears, the crash, the beeping sound of victory and the sound of defeat. Klose’s one sheet of paper is in the same collection.
Auto Race is not a good game by any standard anyone would apply today, and it was barely a good game in 1976. It is something better than that. It is the moment somebody proved a game did not need a screen, a cabinet, a processor or a mains lead, and the fifty years of portable gaming that followed are all just that argument being made at higher resolution.
