Care for a game of Los Alamos Chess?
First, let’s get to the rules.
- Los Alamos Chess is played on a 6x6 board.
- Bishops are removed (the game is also called anti-clerical chess).
- No pawn double-step moves are allowed.
- No en passant moves allowed in Los Alamos Chess.
- Pawns cannot be promoted to bishops.
- No castling.
See, Los Alamos Chess is a simplified game, compared to the classic 8x8 form. It might even be easy to play. That is, unless the year is 1956 and you are the Laboratory’s half-ton computer, MANIAC. If that’s the case, you might struggle a bit.
In an article titled “Experiments in Chess on Electronic Computing Machines,” that appeared in an issue of Chess Review the following year, Los Alamos physicists and mathematicians — and chess enthusiasts — Paul Stein and Stan Ulam detailed their recent results from three Los Alamos Chess games played at then-called Los Alamos Scientific Laboratory. Each game was played by MANIAC (Mathematical Analyzer, Numerical Integrator and Computer), which the article described as a “high speed digital computer” that had “played an important part in thermonuclear calculations.”
The 2 principles
Stein and Ulam asserted that two principles were fundamental in determining “best play” for a chess move: material advantage (getting and keeping pieces) and superior mobility (creating more freedom to maneuver and limiting the same for the opponent). “No one can doubt the validity of these elementary desiderata,” they wrote. “(T)he interesting question is how far method of play can be based on these two principles alone.”
To test this, Stein and Ulam, along with their colleagues James Kister, William Walden and Mark Wells, “decided to construct a method (technically known as a ‘code’), which would enable an electronic computing machine to play chess utilizing just these two criteria of material advantage and mobility.” In their code, the researchers defined mobility “as the number of legally available squares,” and material strength “was evaluated by assigning to each (p)awn the value of eight legal moves and valuing the other pieces accordingly.”
They were not the first scientists to code a computer to play chess. Stein and Ulam note that Alan Turing had done this previously, but “Turing’s code allowed his electronic player to see only one move ahead,” and was not a successful match for a human player. “The reason we chose this game for our initial experiments,” wrote Stein and Ulam of Los Alamos Chess, “was to enable a machine to look two moves ahead, i.e., two moves by each side, and still make its moves in a reasonable time. Therefore, we get a factor of 16 = 24 for chains, and the machine could make a move in about 10 minutes.” This was high speed computing in the Eisenhower era.
Game on
The scientists set up three matches for MANIAC. The first was MANIAC versus MANIAC. The second was MANIAC vs. Martin Kruskal, who was a mathematician and physicist at Princeton University, as well as a skilled chess player. The third was MANIAC versus a politely unnamed Labbie who received a crash course in chess basics over the span of a week.
The first game produced clunky maneuvers (“White continues to waste time,” “Black failed to see the win”) that keyed in the scientists to needed tweaks in the code, such as adding weighting to the mobility value. Still, in the second game with Kruskal, Stein and Ulam observed that MANIAC’s “shortsightedness has not yet been cured.” Kruskal was able to predict the computer’s moves. With the match running towards its inevitable conclusion, the scientists wrote, “In eminently human fashion, the machine ‘deliberated’ 20 minutes before giving up its Queen.” Kruskal handily defeated MANIAC.
The computer’s time to shine was in the third game. Stein and Ulam considered MANIAC’s maneuvers to be more precise than the opponent’s, whose plays they likened to “aimless wandering.” MANIAC was the victor. “All things considered, the machine handled the final attack most impressively.”
The next move
From the results of their three MANIAC games, the scientists concluded in the overruling importance of material advantage and mobility. Yet they wrote, “we feel that the game cannot be reduced to any mathematical formula — not even as complicated and extensive as Einstein’s tensor equations.” They also asserted that coding a computer to play master chess — not just a rudimentary game — was at least 20 years away (the first computer to gain a master rating, CRAY BLITZ, would do so in 1981 at the Mississippi State Championship). Stein, Ulam and their colleagues summarized their chess findings in an article for the Proceedings of the Western Joint Computer Conference.
Curious about all things computers? Check out National Security Research Center historian Nic Lewis’ article about 70 years of electronic computing, and stop by the Bradbury Science Museum’s supercomputing exhibit to learn more about MANIAC, codes and the Lab’s computing legacy.