A new Monte Carlo algorithm is implemented for simulation of the density of states of free-standing polymer films. The algorithm combines the original idea of conducting a random walk in energy space with advanced trial moves such as configurational bias and end-bridging. Excellent agreement is found between the results of this new method and those from simulations in the canonical ensemble, down to temperatures in the vicinity of the apparent glass transition. The efficiency of the new algorithm is studied as a function of the types of trial moves employed. It is found that, depending on the range of energy and density, certain localized moves fail to converge to the correct distribution of states.