1.The relationship between diversity and ecosystem functioning is often analysed by partitioning the change in species performance in mixtures into a complementarity effect (CE) and a selection effect (SE). There is continuing ambiguity in the literature on the interpretation of these effects, mainly in their relationship to ecological mechanisms and processes. 2.Here, we present the emergence of complementarity and selection effects in the results of an evolutionary game theoretical model for plant competition which is exclusively based on competition for light. Eight plant strategies, differing only in the time of onset of flowering, were played against one another to determine the relationship between plant trait differences and CE vs. SE. 3.We show that competitive exclusion may occur even in the presence of a positive CE, i.e. when CE > 0. CE was highest at intermediate differences in flowering time. Increasing trait differences may, therefore, increase CE without leading to coexistence. 4.SE was strongly dependent on which of the strategies was paired. SE was mostly positive if one of the players was early flowering (with low seed yield) and the other strategy had higher monoculture yields. SE was mostly negative if one of the players was late flowering, leading to low monoculture yield for this strategy, but to high gains when competing with strategies that produce less leaf area. The average SE was negative over all pairs. 5.Synthesis. These results show that CE may be positive if the species interaction is beneficial for only one of the two competitors, and the sign of SE depends critically on the underlying mechanism for performance in monoculture. Positive complementarity may go hand in hand with competitive exclusion, while widely different outcomes in SE are possible within a single mechanistic background. Consequently, the way SE changes with increasing richness are not only related to a sampling/selection effect but also by the way the interacting species affect the competitive environment. Interpreting the additive partitioning method with a set notion in mind about the driving mechanisms could lead to incorrect conclusions on how species richness effects drive ecosystem functioning.