Coleco’s “deliberately bad” Donkey Kong for Atari

If you grew up playing Atari, you probably heard the story. Coleco made a deliberately bad Donkey Kong port for the Atari 2600 so they could sell more Coleco Vision consoles. In this blog post, we’ll explore the two sides of the argument that Coleco tanked its Donkey Kong port to hurt Atari.

What was wrong with Coleco’s Donkey Kong for Atari

I can see why the casual observer would think that Coleco tanked the port, even though the original programmer, Garry Kitchen, vehemently denies it. It’s really only half the arcade game, with only two levels. Coleco’s own version, a launch title for the Coleco Vision console, had three levels and graphics much closer to the arcade.

But on the Atari, the barrels you have to jump over don’t look like barrels, they look like chocolate chip cookies. And Donkey Kong looks like a gingerbread man, not at all like a movie monster villain.

Activision, the legendary 1980s game developer, must have believed Garry Kitchen. They hired him not long after, and working for them, he programmed one of the best games on the console, 1983’s Keystone Kapers.

What Coleco’s Donkey Kong for Atari did well

Let’s talk about why Activision, no slouch when it came to Atari 2600 video games, would have believed Garry Kitchen. That’s because the game does several things very well. The most important is that it doesn’t flicker. When you play Pac-Man on the Atari 2600, you notice the ghosts flicker a lot. But there was no flicker in Donkey Kong. And Mario and Pauline’s images are multicolor. The 2600 didn’t provide hardware support for multicolor sprites like later consoles did, so he had to do some software trickery to make that happen. And the slanted platforms were even more difficult, because on the 2600, you drew the graphics one line at a time. An intentionally bad port would not have used slanted platforms. It made programming the game more difficult.

With the Atari, the programmer draws the screen one line at a time, and has to consume exactly 76 clock cycles per line. Both game logic and the graphics have to fit in those 76 cycles, a technique called racing the beam. Drawing slanted platforms makes that logic more difficult, and leaves fewer cycles for something else. It’s also not possible to save any cycles for later. The program has to consume any leftover time by running NOP instructions.

And when you look at two contemporary knockoffs, King Kong by Tiger Vision and Pac-Kong by Suntek, neither of those games bothered with slanted platforms in the first level.

The argument for a 4K cartridge vs 8K

Garry Kitchen has said that he wanted to make the game 8K in size rather than 4K, or at least that he could have made the game better if he’d had 8K to work with. With 8K, he would have been able to make all four levels and potentially improve the graphics a bit.

Coleco didn’t want to do that, because 8K ROM chips were more expensive. It would have made the game cost more to produce, so either Coleco would have had to settle for lower profit margins or raise the price. The price wouldn’t have been ruinous, but if a 20% increase in production cost didn’t project to lead to at least a 20.1% increase in sales, there was no incentive to do it.

We can only speculate what else went into their analysis. Maybe they didn’t want all four levels in the version for the rival console. Maybe they determined having all four levels wasn’t worth the loss in price flexibility. More likely it was the latter. And I’ve heard more than one Gen Xer admit they couldn’t beat more than two levels of Donkey Kong as a kid anyway. I only personally knew one person who could beat three.

More importantly, as Kitchen said in a 2023 interview, he only had 90 days to program it, and 90 days was only enough for the 4K version. Taking longer would have meant the cartridge wasn’t ready for Christmas 1982.

Why modern Donkey Kong versions are better

There are a pair of modern reimaginations of Donkey Kong for the 2600 that are much closer to the arcade version. But even those have trade-offs. Both of them have very impressive graphics and sound. But one of them doesn’t have slanted platforms. The other one has slanted platforms but is extremely flickery. And they are both 32k in size, eight times what Garry Kitchen had to work with.

Considering the limitations of the console, the tradeoffs necessary to squeeze a recognizable version of Donkey Kong into just 4K of ROM space, and only 90 days of working time to do it, Garry Kitchen made a good game. It would have been interesting to see what he would have built with 8K to work with. But even that would have been disappointing compared to what modern implementations do with modern development tools, no particular deadline to hit, and 32k of ROM space to work with. But a 32k cartridge would have been out of the question in 1982.

The contemporary knockoffs also were 4K cartridges, so they had similar limitations. They made a different set of trade-offs, but I can’t really say either of them were objectively any better.

I remember playing Tigervision’s King Kong in the ’80s at a friend’s house. He didn’t try to pass it off as anything other than a Donkey Kong imitation. It wasn’t better. It wasn’t necessarily worse. Mostly it was just different. Different enough to avoid a Nintendo lawsuit. The platforms weren’t slanted and the giant ape villain still looked more like a gingerbread man. If it had been practical do do much better with 1982’s tools and techniques, Tigervision had the incentive to do it. Mostly their clone just validated Coleco’s decisions.

Did Coleco contribute to the video game crash?

We could argue the 2600 version of Donkey Kong was part of the problem of low-quality games on the 2600, but we could also argue it was a symptom. Arcade games were getting better, and the 2600 wasn’t keeping pace. The 2600 version of Pac-Man was a letdown, and to a similar extent, the 2600 version of Donkey Kong was as well. The market was getting flooded with low quality knockoff games, and even an entire knockoff platform, the Emerson Arcadia 2001.

If the decision makers had set the bar somewhat higher than they did, then maybe the 1983 video game crash wouldn’t have happened. But it would have taken more than one publisher making that decision. And they weren’t worried about the ecosystem. Their only incentive was the current quarter’s profits.

You can’t exactly expect multiple publishers to do that on their own. And that is why Nintendo and other future consoles didn’t allow third party development without approval from the console maker.

So it would have been nice if Coleco had allowed Garry Kitchen to make an 8K version of Donkey Kong. But in the long run, it probably wouldn’t have made much of a difference. And the idea that Coleco would deliberately release low quality games for a rival console doesn’t make sense economically. In 1982, Coleco sold $8 million worth of game cartridges. Only 2 million dollars of that was for its own platform. The majority of those sales were Atari carts.

Coleco’s Donkey Kong for the Atari 2600 wasn’t a great game. But it also wasn’t nearly as bad as it could have been. Looking at it today, it makes it easy to ask why we liked Atari when better versions of most of the games existed on other consoles. But in 1982, it was a good-enough version of the year’s hottest game for a console that millions of people already had. That’s why Coleco’s Donkey Kong is one of the most common cartridges for the Atari 2600.