In the paper we investigate the role played by both production and market risks on farmer’s decision to adopt long rotations (over 2 years), considered as innovative cropping systems. We build a multiperiod dynamic farm model (run under GAMS) that arbitrates each year between traditional and innovative rotations. With discrete stochastic programming, the production risk is accounted as an intra-year risk; yearly farming operations are declined according to a decision tree where probabilities are assigned. Subjective yield and cost distributions linked to this decision tree are elicited among a sample of 13 farmers that are experiencing this innovation in South-western France. The price risk is randomly distributed with a given market trend. The crop acreage can be revised according to the market situation. The simulations show that substantive sunk costs are incentive to remain in the long rotation when the farmer is already engaged and when he is supported for this engagement. They also show that both a high risk aversion and a highly positive market trend tend to slow down the conversion towards innovative systems.