Predicting expert–novice performance as serious games analytics with objective-oriented and navigational action sequences

New metrics are needed to produce insights in serious games analytics.Action sequences can be coded using task- or tile-based approach.Size of map grids can affects the accuracy in performance prediction.Grid size affects tile-based coding of action sequences and performance prediction.The optimal map grid size must be determined independently for each game. Previous research differentiated expert vs. novice performances based on how (dis)similar novices' action sequences were from that of the expert's by way of similarity measures. Action sequences were coded using an 'objective-oriented' (or task-based) approach based on the sequence of objectives/tasks completed in-game. Findings from these studies suggest that the task-based similarity measures is a better predictor than (a) distance traversed, and (b) time (of completion).In this study, we suggested an alternative method to code action sequences of experts and novices by way of a 'navigational' (or tile-based) approach. We divided a game-map into grids/tiles of different sizes to facilitate tracing of the path traversed by players in game and proceeded to test the effect of grid sizes on differentiating between experts and novices. We further compared the two different action sequence coding approaches and their abilities to measure players' competency improvement in serious games. The results of the study showed that the size of game grids does matter, and that both task-based and tile-based action sequence coding approaches are useful for serious games analytics.