How Atari Made Chess Possible on 128 Bytes (for non-techies)
Video Chess: When 128 Bytes Was All You Had
I grew up in the 1980s, back when video games were in their infancy. Our first gaming system was Pong, which was a short-lived novelty. Then my dad brought home the most coveted console on the market: the Atari 2600!
The 2600 was released in 1977 for $199. That sounds reasonable today, but adjusted for inflation, it's the equivalent of over $900! I still have no idea how my family afforded it, especially since we didn't have much money at the time.
Look at this beauty, complete with simulated wood that matched the simulated wood on our station wagon:
One of the many games available for the system was Video Chess, released in November 1979 for a whopping $40—about $170 in today’s dollars. My dad was more into shooter games, so it was never part of our collection. Since I'd never played it, I recently decided to try it out on an Atari 2600 emulator. It was simultaneously painful and impressive.
A Look at the Game
Here's what the screen looks like:
If you're wondering why the pieces look strange, you're not alone. The game didn't have enough memory to display the full chess pieces all at once. The developers used a technique called "venetian blinds," where alternating rows contained the pixels that made up the piece's graphic. Take a look at this zoomed-in screenshot to see it up close:
This workaround was necessary because the Atari 2600 had only 128 bytes of RAM. To put that in perspective, an email saying "hello" is probably larger than 128 bytes! Even a short tweet on X wouldn't have enough memory to be stored on the 2600. Our modern smartphones have billions of bytes of memory, so the fact that the 2600 could do anything at all with 128 bytes is remarkable.
The Genius Behind the Code
Because of these limitations, many programmers at the time felt a chess game on the 2600 was impossible. So how did they do it?
With so little memory, the 2600 could only "remember" the current state of the board. When you made a legal move, the screen would flash as the game updated its memory, and the computer would begin to "think."
This "thinking" involved building a tree of possible moves. The program would evaluate its own moves and potential human responses, using a scoring system to find the move that gave the computer the highest score and the human the lowest. It would even prune the tree by cutting off branches of obviously bad moves, such as hanging the queen, to free up memory for other options.
The difficulty setting you chose at the start of the match determined how many moves ahead the computer would consider, which is why a high-level game could take several hours for the computer to find its next move.
Once it found the best move, the program would update the board, display it for the human player, and the game continued. The fact that Video Chess could even support moves like castling and en passant, which were uncommon for computerized chess at the time, is a true testament to the ingenuity of the developers. Castling seems simple today, but the program had to consider all the pieces on the board to ensure the king didn't pass through or land on a checked square, which is incredibly impressive with only 128 bytes of memory.
My Humble Defeat
So how did my game against the 2600 go? Honestly, I ran out of patience. The wonky appearance of the pieces was so irritating that I was constantly asking myself, "Is that the bishop? The queen... or?" Eventually, I had to stop.
Despite its limitations, Video Chess still has a surprising reputation. It's both humbling and humorous that modern AI engines like ChatGPT and Copilot have been played against this ancient chess engine on its easiest setting, and the 2600 has actually come out on top!
So, a huge congratulations to the original programmers, Larry Wagner and Bob Whitehead. Your creation was truly an impressive achievement for its time.
