Los Alamos Chess
Early work on computer chess at Los Alamos used a simplified 6×6 variant without bishops or castling so that the MANIAC I machine could handle the game’s combinatorial complexity, leading to one of the first recorded wins by a computer over a human in a chess-like game. Commenters explore why specific rules (like banning bishop promotion) were chosen, how board size and piece sets affect search space and solvability, and how this compares to modern engines and variants such as 5×5 “Gardner” chess. The conversation branches into broader reflections on early hardware constraints, naming and classification of game variants, and recommendations for faster, more approachable chess-like games.
Chess variant design & rules
- Los Alamos chess removes bishops entirely and forbids promotion to bishops to avoid implementing extra piece logic on a highly constrained machine.
- Retaining knights is defended: they generate fewer moves on average than bishops, especially on a small board, reducing branching factor while preserving interesting tactics.
- Queens stay in; in one historical game the human played without a queen to handicap in favor of the computer.
- Smaller boards and reduced pieces are seen as making games faster, easier to pick up, and still strategically rich.
Computational limits & solving complexity
- Thread debate on whether the variant could be “solved”:
- Full tablebases beyond 7 pieces in normal chess are already enormous; extending that logic suggests 24-piece Los Alamos chess is still intractable for complete tablebases.
- Others note that “weak solving” (proving the game-theoretic value from the starting position) is more feasible, citing a 5×5 mini-chess that has been weakly solved without tablebases.
- Branching factor, board size, and reduced rules (no castling, en passant, double pawn push) significantly cut search space, but not enough for full perfect-play databases.
Historical computing & naming
- MANIAC’s 20-minute move times prompt reflection on how hard early coding and debugging were (e.g., tracing programs via card layouts on the floor).
- A long technical correction distinguishes slow relay machines from much faster vacuum-tube computers, arguing:
- Tube-based machines already did tens of thousands of operations per second.
- Core memory latency, not tube speed, was often the bottleneck.
- Architectural advances and miniaturization, more than raw transistor speed, drove the jump from kiloflops to today’s tera/petaflops.
- Nostalgic discussion of old machine names (MANIAC, Colossus, etc.) versus modern “corporate” names.
Other games and variants
- Multiple recommendations for short, chess-adjacent modern games, especially Onitama (5×5, move cards), plus mentions of Hive and Santorini.
- Link to an online Los Alamos chess implementation; noted bug where pawns incorrectly promote to queens.
Sudoku-variant naming analogy
- Lively side debate compares “sudoku variants” with chess variants:
- Some dislike puzzles marketed as standard sudoku that rely on extra, sometimes arbitrary rules optimized for spectacle or computer search.
- Others argue variants are legitimate extensions, analogous to chess families; they explore new constraint interactions and can still feel like “the same” logical game.
- Disagreement over naming: whether such puzzles should still be called sudoku or something entirely different.
Miscellaneous observations
- Clarifications that Los Alamos chess was not literally the first time a machine beat a human in a chess-like setting; earlier special-purpose automata existed.
- Casual players often don’t know en passant; it’s a frequent source of confusion and online memes.
- Several commenters jokingly expected a nuclear-themed ruleset from the name and sketch out hypothetical “atomic” mechanics involving piece decay or pawn “launches.”