Recruitment forecasting constitutes a major issue in population dynamics, especially in stock assessments. In many cases, high recruitment stochasticity thwarts the determination of a stock-recruitment (SR) relationship. When no SR relationship is available, recruitments can be projected from past trends using methods such as the geometric means of the past recruitments. This approach implicitly assumes stability in upcoming years and might not be appropriate in the case of regime shifts. Recruitment forecasting is particularly critical when predicting fishing opportunities for recruitment fisheries. The fishery of glass European eels (Anguilla anguilla) is an example of a recruitment fishery. The French glass eel fishery, which is currently guided by two models to forecast upcoming recruitments and derive fishing opportunities, has failed to achieve management targets. This study develops new models that forecast glass eel recruitments without assumptions about the SR relationship, and with flexible assumptions about future recruitment evolutions. Tests based on multiple criteria found the best performance in a random slope model, which provides flexibility to track recruitment dynamics while balancing the interests of fisheries, management and conservation. This model has wide potential for trend projections in natural resource management, economics and other fields.