A simple and exhaustive model evaluation technique for regional chemistry-transport is discussed. It is based on the concept that we can learn more on models performances by comparing the results to in situ measurements available for other time periods than the period originally targeted in the simulation. First, the usual scores (spatial and temporal correlation) are computed for a reference period, using the actual temporal synchronization and spatial location of measurements . Second, the same scores are calculated for several other years by conserving only the actual spatial locations and days of the year. The difference between the two score provides complementary insights to the following questions: (i) is the model performing well only because the situation is persistent? (ii) is the model representative enough of the measurements for all variables? (iii) if the pollutants concentrations are not well modelled, is it due to meteorology or chemistry? In order to synthetize the large amount of results, a new score is proposed: the "multi-year variability", designed to compare the several indicators between all the years of validation and to quantify if the studied period was well modelled and, if yes, for the good reasons.