This paper presents an attempt to construct a ”practical” CSP algorithm that assigns a variable with 2 values at every step. Such a strategy has been successfully used for construction of ”theoretical” constraint solvers because it decreases twice the base of the exponent of the upper bound of the search algorithm. We present a solver based on the strategy. The pruning mechanism of the algorithm resembles Forward Checking (FC), therefore we term it 2FC. According to our experimental evaluation, 2FC outperforms FC on graph coloring problems and on non-dense instances of randomly generated CSPs.