We combine two known polynomial time approximation algorithms for the maxim
um traveling salesman problem to obtain a randomized algorithm which output
s a solution with expected value of at least r. times the optimal one for a
ny given r < 25/33. (C) 2000 Elsevier Science B.V. All rights reserved.