Soil moisture anomalies underpin a number of critical hydrological phenomena with socioeconomic consequences, yet systematic studies of soil moisture predictability are limited. Here, we use a data-adaptive technique, Linear Inverse Modeling, which has proved useful as an indication of predictability in other fields, to investigate the predictability of soil moisture in northern California. This approach yields a model of soil moisture at 10 stations in the region, with results that indicate the possibility of skillful forecasts at each for lead times of 1–2 weeks. An important advantage of this model is the a priori identification of forecasts of opportunity—conditions under which the model’s forecasts may be expected to have particularly high skill. Given that forecast errors (and inversely, their skill) can be estimated in advance, these findings have the potential to greatly increase the utility of soil moisture forecasts for practical applications including drought and flood forecasting.