Zero-mean unbiasing as a preprocessing step for greedy TSP solvers improves the quality of solutions on average.