The Science of Rock Paper Scissors: Research, Studies & Discoveries

Researchers use Rock Paper Scissors to study mixed strategies, conditional responses, and non-transitive competition. A 2014 experiment followed 360 students in sixty six-person groups for 300 rounds; a separate 2016 experiment tested 31 undergraduates against an equal-random computer. Both report protocol-specific behavior, not a universal way to predict a person. Ecological papers use “rock-paper-scissors” as an analogy for cyclic competition, while the University of Tokyo robot wins by reacting after motion begins rather than forecasting a choice.

Yes, This Is a Real Field of Study

Rock Paper Scissors gives researchers a compact three-action game with a clear mixed-strategy baseline. The papers reviewed here ask different questions using different populations and protocols, so their findings should not be combined into a single claim that all people behave the same way.

The Zhejiang University Study (2014)

Wang, Xu, and Zhou assigned 360 Zhejiang University students to sixty groups of six. Participants played 300 rounds with random pairings inside their group and cash incentives tied to results.

After wins: The reported conditional-response measurements and model included a tendency to repeat an action after winning.
After losses: Aggregate transition probabilities were consistent with cyclic shifts in the study’s population.
After draws: The paper reports outcome-conditioned transitions; it does not attribute them to boredom or another unmeasured emotion.

The authors used conditional response to explain population-level cycling despite near-equal aggregate action frequencies. That result belongs to this experiment and does not reveal any particular opponent’s next throw without later, opponent-specific validation.

Wang, Xu & Zhou (2014), Scientific Reports, doi:10.1038/srep05830

Evolutionary Biology

Biologists also use the RPS cycle as a mathematical description of non-transitive competition. These organisms are not playing the hand game:

Side-blotched lizards: Sinervo and Lively followed three male morph strategies over six years and described frequency-dependent cycles using an RPS model.
Microbial system: Kerr and colleagues tested toxin-producing, toxin-sensitive, and toxin-resistant bacterial strains. Local dispersal helped maintain diversity in their experimental system.

These studies show that cyclic dominance can help explain coexistence under stated ecological conditions. They do not establish a universal law of nature or evidence about human hand-game behavior.

Artificial Intelligence

The best documented machine on this page demonstrates the difference between prediction and reaction:

University of Tokyo Janken robot: A high-speed vision system recognizes the developing human hand and produces the winning gesture. The laboratory describes a 100% winning rate; because the system responds after motion begins, it is a reaction demonstration rather than evidence that a model predicted an unrevealed choice.

University of Tokyo Ishikawa Group: Janken robot project

Neuroscience

Kikuchi and colleagues used near-infrared spectroscopy in a modified win-versus-lose RPS task with 15 healthy volunteers. The paper reports prefrontal measurements for that task; it does not show that ordinary throws can be decoded from body language or that every RPS decision activates a fixed list of brain regions.

Kikuchi et al. (2007): prefrontal activity during a 15-person RPS task. This study tested a specific win-versus-lose task and does not establish that every ordinary throw can be read from brain activity or body language.

Use in education: The cited experiments can support lessons about experimental design, sample limits, probability, and game theory when students keep the hand game separate from ecological analogies.

The Zhejiang University Study (2014)

Wang, Xu, and Zhou assigned 360 Zhejiang University students to sixty groups of six. Participants played 300 rounds with random pairings inside their group and cash incentives tied to results.

After wins: The reported conditional-response measurements and model included a tendency to repeat an action after winning.

After losses: Aggregate transition probabilities were consistent with cyclic shifts in the study’s population.

After draws: The paper reports outcome-conditioned transitions; it does not attribute them to boredom or another unmeasured emotion.

Evolutionary Biology

Biologists also use the RPS cycle as a mathematical description of non-transitive competition. These organisms are not playing the hand game:

Side-blotched lizards: Sinervo and Lively followed three male morph strategies over six years and described frequency-dependent cycles using an RPS model.

Microbial system: Kerr and colleagues tested toxin-producing, toxin-sensitive, and toxin-resistant bacterial strains. Local dispersal helped maintain diversity in their experimental system.

These studies show that cyclic dominance can help explain coexistence under stated ecological conditions. They do not establish a universal law of nature or evidence about human hand-game behavior.

Artificial Intelligence

The best documented machine on this page demonstrates the difference between prediction and reaction:

University of Tokyo Janken robot: A high-speed vision system recognizes the developing human hand and produces the winning gesture. The laboratory describes a 100% winning rate; because the system responds after motion begins, it is a reaction demonstration rather than evidence that a model predicted an unrevealed choice.

Neuroscience