This risk study had tried to add a new dimension to the existing risk analysis. Taking advantage of the analytical model that had been developed by Zandstra (1979) which in short is what the probability the realized production is below break event production, the study had tried to use it to estimate or measure the production risk of soybean farming in Central Java. Normality assumption of productivity distribution was tested using crop cut figures of 1990-1992. A survey data of soybean farming in irrigated lowland, rainfed lowland and upland of the same period was used to calculate the break even production of respective ecosystem. Results of the analysis showed that production risk estimate of soybean farming in irrigated lowland is Rp 521.000 each hectare. There are variations among planting season and the lowest risk is in the Januari-April period reaching about 46 percent of total cost. In the rainfed lowland has an average of nominal risk of Rp 368.000 each hectare and the Mei-August planting shows smallest risk. In the upland average nominal risk is Rp 358.000 each hectare and the lowest risk exists in the September-Desember planting season. However, risk indicator of risk/each kg of product is more suitable or in line with lost or profitability of the soybean farming in the season. Making more clear categorization of different lowland ecosystem of crop cut data was suggested.