Abstract: A new optimization method, contraction-expansion algorithm, is developed and applied to fit nonlinear equations. Five points with equal distance are allocated in the given initial intervals, then the algorithm searches for the points of better objective function one round after another, with contracting and expanding intervals alternatively.The step length (l), a critical factor to optimal solution, is determined and adjusted by feedback information of the searching process. The experiments with number of simulated and real data sets show that the new algorithm is much easier to the optimization than other methods. The algorithm does not need to provide derivatives of the equation.This algorithm can also be used in other optimization problems.