Wednesday, November 18, 2026Wed, Nov 18

Moon: waxing gibbous, 58% litAurora: active

Español

Space, Earth and the natural world, explained.

Chess Engines Keep Board Tension High After Move 20, Humans Ease Off

Human chess players lower strategic tension after 20 moves while engines like Stockfish keep it high, a study in APS Open Science finds.

Strategic close-up of chess pieces on a board.
Photo: Theia Sight / Pexels

Human chess players tend to calm the board down after about 20 moves each, while top chess engines keep it tangled, according to a study published in APS Open Science and reported by Scientific American. The finding points to one habit that may help explain why machines have beaten people at chess since Deep Blue defeated the world’s best player in 1997.

The researchers measured what they call strategic tension: the number of possible interactions between opposing pieces at a given moment. Tension usually rises when more moves are available. In games between people it peaked early and then fell, but in games between engines it stayed high.

How the team measured tension

The study’s senior author is physicist Vito D. P. Servedio of Complexity Science Hub in Vienna. His group pulled games from online databases of two kinds: human against human, and the engine Stockfish against the engine Leela. They then scored the tension at every point in each game.

Both engines belong to the modern generation of chess software. Early programs such as Deep Blue used tree search, which means playing out many branches of possible future moves internally before choosing one. Stockfish and Leela also search trees, but trained neural networks guide which branches they explore.

The study is observational. It compares patterns in recorded games and does not test what happens when a player is pushed to keep tension high.

What changed at move 20

Up to the point where each side had made 20 moves, human games carried more average tension than the engine games. People in the opening often lean on memorized lines and hope a complicated position will trip up a human opponent. After that point, tension in human games dropped below the engine level, and the engine games stayed tense.

One explanation is that people simplify the position once it becomes unfamiliar, because they cannot track many interacting pieces and options in their heads at once. Merim Bilalić, a psychologist at Northumbria University in England who was not part of the study, said the result makes sense. Humans need to lower their mental load, he told Scientific American, while “engines don’t really have that problem.”

Stronger players and longer games ran hotter

Higher-rated human players kept more tension on the board throughout their games. Time controls mattered too: tension was highest in classical chess, lower in rapid chess and lowest in blitz, the fastest format. When Stockfish was set to search deeper trees of possible moves, the tension in its games went up.

Tension tracked skill in both people and machines, but the study did not show which way the effect runs. It is unclear whether high tension produces better play, or whether stronger players and deeper searches simply create more of it.

Why it matters beyond chess

Researchers still use chess as a model for studying how humans and machines think. Servedio said his group’s paper is probably the closest anyone has come to explaining what the engines are doing. Without that, he said, “it looks like they are playing magic.”

He also suggested the idea could carry over to higher-stakes settings such as diplomacy, where an AI might reach better outcomes while making a situation more complicated. Servedio warned that systems under high tension are unstable, and that a small error by either side could then lead to disaster. That application is his view and was not tested in the study, which looked only at chess games.

Source: Scientific American