# Non-Transitive Games
**Entity class:** Game-theoretic mechanism
**Domain:** Evolutionary game theory / multi-agent learning
**Maturity:** Developed
## Definition
A **non-transitive game** contains a cycle of advantage: one strategy defeats another, which defeats a third, which in turn defeats the first. [[wiki/Game Theory|Rock, Paper, Scissors]] is the minimal example. Because no move dominates all others, the value of a strategy depends on the frequencies and responses of other players; in the symmetric, zero-sum, simultaneous version, the unexploitable mixed strategy assigns equal probability to each move. That equilibrium describes strategic incentives, not a guarantee that every evolving population will maintain all three strategies.
## Population structure and learning
In side-blotched lizards, three male mating strategies form a frequency-dependent cycle. In Kerr and colleagues' bacterial experiment, toxin production, toxin resistance and sensitivity also formed a cycle, but local interaction preserved coexistence where a well-mixed environment lost it. Reichenbach, Mobilia and Frey's spatial model likewise found that mobility can support or destroy diversity depending on whether it crosses a critical regime. **The cycle creates an opportunity for coexistence; spatial structure and dynamics determine whether coexistence lasts.**
Repeated play adds a second timescale: each participant changes strategy in response to others, so the population's payoff landscape changes while it is being learned. Human Rock, Paper, Scissors experiments found conditional response and population-level cycling, while [[wiki/Autocurriculum|autocurricula]] and agent leagues deliberately use diverse opponents to expose policies that overfit one rival. [[wiki/Replicator Dynamics|Replicator Dynamics]] tracks changing strategy frequencies; [[wiki/Nash Equilibrium|Nash Equilibrium]] describes a different, static stability question. [[articles/Rock, Paper, Scissors|Rock, Paper, Scissors]] develops the route from the child's game through ecology, machine learning and institutions.
## Sources
- [Sinervo and Lively, *Nature* (1996)](https://doi.org/10.1038/380240a0) — cyclic mating strategies in side-blotched lizards.
- [Kerr et al., *Nature* (2002)](https://doi.org/10.1038/nature00823) — bacterial competition under local and mixed dispersal.
- [Reichenbach, Mobilia and Frey, *Nature* (2007)](https://doi.org/10.1038/nature06095) — mobility threshold in a spatial model.
- [Wang, Xu and Zhou, *Scientific Reports* (2014)](https://pmc.ncbi.nlm.nih.gov/articles/PMC5376050/) — conditional human responses and social cycling.