Interannual variability of seed production, known as masting, has far-reaching ecological impacts including effects on forest regeneration and the population dynamics of seed consumers. Because the relative timing of management and conservation efforts in ecosystems dominated by masting species often determines their success, there is a need to study masting mechanisms and develop forecasting tools for seed production. Here, we aim to establish seed production forecasting as a new branch of the discipline. We evaluate the predictive capabilities of three models - \emph{foreMast}, $\Delta$T, and a sequential model - designed to predict seed production in trees using a pan-European dataset of \emph{Fagus sylvatica} seed production. The models are moderately successful in recreating seed production dynamics. The availability of high-quality data on prior seed production improved the sequential model's predictive power, suggesting that effective seed production monitoring methods are crucial for creating forecasting tools. In terms of extreme events, the models are better at predicting crop failures than bumper crops, likely because the factors preventing seed production are better understood than the processes leading to large reproductive events. We summarize the current challenges and provide a roadmap to help advance the discipline and encourage the further development of mast forecasting.