Where do mistakes lead? A survey of games with incompetent players (Q6159517)

The paper under review is to survey relatively recent research on predicting assumption that player incompetence is real and ubiquitous. Player's incompetence is referring to unable to execute his/her chosen strategy due to a lack of skill, and inherently complicates a game. The key conceptual assumption: [A1] Incompetence manifests itself as a set of probability distributions on sets of actions available to one or more players. So far the topic of incompetent game evolves along two distinct directions, one for the study of classical non-cooperative games with at least one player incompetent and the other for the study of evolutionary incompetent games. Section 2 surveys the classical non-cooperative games with incompetent started from the original incompetent bimatrix game by \textit{J. D. Beck} et al. [Automatica 48, No. 10, 2400--2408 (2012; Zbl 1271.91009)], where the distribution is represented by the incompetent matrices for player 1 and player 2. And serves an essential building block to construct dynamic learning games. Attack-defence game with incompetence is given to have the bimatrix game with connection between complete mixedness and the game value plateau. Theorem 1 shows that there is a natural one-to-one correspondence between equilibrium in the competence game and equilibrium in the incompetent game under the incompetent bimatrix, and the equilibrium lies in the interior of executing strategies if the incompetent bimatrix is weakly completely mixed. Theorem 2 shows the equilibrium of the incompetent bimatrix game is convergent under the limits of incompetent matrices, and Beck et al. [loc. cit.] study the dependence of equilibrium-induced expected rewards on the players' learning parameters, that is presented in Theorem 3 and Theorem 4 for different conditions on incompetent bimatrices. The incremental learning game is a stochastic game unfolding over an infinite time horizon between repeated players of an incompetent game. Theorem 5 gives a sufficient condition (simplified version of the equilibrium inequalities) to have strategy profile as an equilibrium of the stochastic game. Section 3 surveys incompetence in biological populations. A mixed strategy is an evolutionary stable strategy (ESS) if the frequency-dependent fitness of the entire population is bigger than the fitness of a population adopting strategy in a population of individuals using a strategy, and the ESS is a special case of a Nash equilibrium. \textit{M. Kleshnina} et al. [``Prioritised learning in snowdrift-type games'', Mathematics 8, No. 11, Article ID 1945, 20 p. (2020; \url{doi:10.3390/math8111945}); ``Shifts in evolutionary balance of microbial phenotypes under environmental changes'', R. Soc. Open Sci. 9, No. 11, Article ID 220744, 17 p. (2022; \url{doi:10.1098/rsos.220744)}] call the bifurcation point of the replicator dynamics a critical value of the incompetent parameter and prove that the incompetent game possesses an ESS if the fitness matrix possesses an ESS, and the strategies are convergent under the incompetent parameters approach. \textit{S. A. Levin} [Bull. Am. Math. Soc., New Ser. 40, No. 1, 3--19 (2003; Zbl 1015.92001)] shows that the mixed strategy with an explicit formula is an interior ESS for the incompetent stochastic game and can be determined under the interior ESS mixed strategy for the original fitness game, given in Theorem 7. Theorem 8 gives the vertex as stable point of the replicator dynamics with execution error for edge and vertex related to the fitness matrix. Specifically, if the original game possesses a stable equilibrium, then the solution of the incompetent game with periodic form of incompetence will converge to a stable periodic orbit around this stable equilibrium. Evolutionary game theory has been widely applied to study the evolution of microbes. \textit{E. C. Zeeman} [Lect. Notes Math. 819, 471--497 (1980; Zbl 0448.92015)] proposed that nonmotile strategy dominates, chemotactic strategy dominates, a stable mixed equilibrium exists and an unstable mixed equilibrium exists by the cost of swimming \(c> (<) 1/2\) and \(m > (<) c\). The stochastic learning leads to a situation where most bacteria in a population are able to perform chemotaxis if a chemotactic strategy was dominating. Kleshnina et al. [loc. cit.] show that the sign of the relative strategical advantage fully determines which strategy has to be learnt first, the last example on bacterial motility game and prioritized learning is presented in this section. Section 4 concludes the survey and future extensions for classical games (continuum of action with a general incompetence-adjusted utility function would be in a Riemann-Stieltjes integral, incompetence dependence dependent action spaces, extensions to stochastic games and the extension to incremental learning) and for evolutionary games (generalization of population dynamics, generalizing prioritized learning by exploiting the power of simulations, learning as a function of frequency of strategies).