An evolutionary game theoretical model shows the limitations of the additive partitioning method for interpreting biodiversity experiments