Study Finds Humans Hold Chess Tension Longer Than AI
Researchers measured strategic tension in human and AI chess games, finding human play peaks early before falling below AI levels.
Updated on Nov. 15, 2026 in Artificial Intelligence
Live Poll
Do you think AI should be trusted to help make high-stakes diplomatic decisions?
A study published in APS Open Science compared strategic tension in human and AI chess games, collected from online databases. Average tension in human games exceeded that of AI games after each player made 20 moves, though it then dropped below AI levels, which stayed high.
Why it matters
The findings suggest people may simplify their piece arrangements because they cannot hold complicated dynamics and multiple options in mind at once. The researchers note the results may also bear on higher-stakes settings such as diplomacy, where AI could reach better outcomes but at the risk of complicating matters.
Strategic tension was defined as the number of possible mutual interactions between opposing pieces, and it rose when Stockfish searched deeper trees of possible gameplay. Higher tension correlated with greater ability in both humans and AI, though the study did not specify the exact search depths used.
The players
Vito D. P. Servedio
Vito D. P. Servedio is a physicist at Complexity Science Hub in Vienna who served as the senior author of the study on strategic tension in chess.
Complexity Science Hub
Complexity Science Hub is a research institution in Vienna that studies complex systems across economics, health, and society.
Deep Blue
Deep Blue was an IBM chess computer that defeated the best human player of its time in 1997, and it searched many branches of a decision tree before each move.
The details
The senior author, Vito D. P. Servedio, a physicist at Complexity Science Hub in Vienna, led the work. Higher-rated human players kept greater tension throughout their games, and tension was greater in games with more allotted time.
Timeline
1997: Deep Blue beat the best human chess player at the time.
After each player's 20th move: Human tension drops below AI tension in games.
The Tech Race
Early engines such as Deep Blue won by internally playing out many branches of a decision tree, while Stockfish and Leela now use trained neural networks to guide their search. This study shows the two generations and human players differ in how they manage strategic tension across a game.
For everyday users of chess software, the findings suggest that engines with neural-network search may play in ways that differ from human intuition, which can matter for training and analysis. The study does not address any direct effect on consumer products or pricing.
The takeaway
Human players tend to simplify positions sooner than AI does, which suggests that people manage complexity by narrowing their options. Chess players who study engine games may find it useful to notice where their own tension drops off.
Further reading
Explore more on Artificial Intelligence to see how machine decision-making is shaping games and beyond.
Source note: This article includes information reported by Scientific American.
Live Poll
Do you think AI should be trusted to help make high-stakes diplomatic decisions?
