January 27, 2026
Queens, memes, and infinity
How many chess games are possible?
Is chess bigger than the universe? The comments explode
TLDR: A new estimate suggests “typical” short chess games could number around 10^166, with evidence from sampling methods. The comments erupt over whether chess is effectively infinite, when novelty appears in a game, and if it really beats “atoms in the universe,” turning math into meme-fueled debate.
The article tries to count how many chess games are possible, starting with a simple Fermi-style guess (average choices per move times average moves) and landing on a brain-bending estimate of roughly a 1 with 166 zeros. It even nods to wild “monster” games with too many queens and astronomically large bounds. But the real show is the comments. One camp screams “Infinite!” like it’s a plot twist, while math-minded readers nitpick the assumptions and demand finer-grained proofs. A Go researcher drops a flex, pointing to a “googolplex of Go games,” and instantly the chess vs Go rivalry lights up. Curious fans ask the coolest question of the day: how many moves until you reach a board no human has ever seen? Meanwhile, movie math crashes the party with the classic “more positions than atoms in the universe” line, and the thread erupts into memes about pawns outnumbering protons (video). The author’s switch to a sampling method (Knuth’s path product) gets applause from practical folks, but skeptics still grumble “meh, be more rigorous.” Verdict: it’s math-meets-meme, with a community split between awe, nitpicks, and existential chess jokes.
Key Points
- •The article examines how to estimate the number of distinct chess games, focusing on short games rather than extreme, pathologically long cases.
- •It references François Labelle’s constructions of very long games and cites bounds for long-game counts on the order of 10^29241 to 10^34082.
- •A Fermi-style estimate sets typical game length at about 100 half-moves (based on Thomas Ahle’s data) and uses a sample position with 46 legal moves to estimate ≈10^166 typical games.
- •To reduce subjectivity, the article defines “short” games as those ending within 100 half-moves under rules implemented in python-chess and proposes uniform random sampling of legal moves.
- •It replaces the Fermi estimate with Knuth’s path product estimator computed along sampled game paths, noting it matches the Fermi estimate only if the sampled game is exactly typical.