The search for environmental descriptors of fish stock production is ongoing. Although numerous correlations between environmental variables such as food abundance or sea surface temperature have been proposed in the past, few are operational in a fisheries management context today. Reasons for this may include many -but spurious -correlations, the use of higher-level climate-change indicators, and misconceived perception of causal relationships. In the present study, we demonstrate how modelled oceanographic data describing local conditions, combined with a simple probabilistic risk assessment, can be used to forecast fish recruitment. We used the lesser sandeel Ammodytes marinus in the North Sea as an example, and focussed on the circulation patterns experienced by the first-feeding larvae on the Dogger Bank. A strong link between the net direction of water transport at the surface and unusually strong year-classes of sandeel was found. For example, the most extreme recruitments only took place in years with a particular type of flow regime in February, which may be associated with the occasional reversals of the North Sea circulation. Using risk ratios, we show the potential for using flow regime in probabilistic short-term forecasts of unusually strong year-classes. Lastly, we propose a hypothesis for recruitment in sandeel which could be extended to other species, and thereby contribute to future studies of predictors in recruitment forecasting.