The recent progresses to predict climate fluctuations several months in advance by using statistical or dynamical models, suggest opportunity to improve the management of climatic risk in agriculture. The aim of this paper is to evaluate the potential benetits farmers can gain from seasonal predictions. To do so, we developed a simple bioeconomic model of a typical community in the semi-arid areas of Senegal. The model is used to simulate the decisions of farmers with a priori information on the quality of the rainy season. We first evaluate academically the potential benefits of a perfect prediction scheme then show how these benefits are affected by the decrease of the skill. Then, we evaluate the potential benetits of several widely-used prediction schemes: a statistical scheme based on the relationships between sea surface temperatures and rainfall and the dynamical predictions of DEMETER. The results show that predicting dry seasons is much more useful than predicting the wet ones. lndeed, on one hand the adoption of a wet forecast exposes the farmers to a high risk of failure in case of imperfect forecast by favoring cash crops such as maize and groundnut that are highly vulnerable to drought. On the other hand, the response to a dry forecast minimizes the climate risk by favoring robust crops such as millet and sorghum. We found a slightly similar skill and economic value of the statistical and dynamical forecast methods by considering the saine leadtime and the same 3 1-years hindeast period 1970-1000. The two methods would have been beneficial for farmers over this 31-years period while using only the dry years forecasts with an increase of income of 13.75%, for the statistical model and 9.58% for the bias-corrected DEMETER multi-model-ensemble mean.