Yearly winning percentages of 23 professional basketball teams over a 10-year period were used to evaluate the stability of team performance. The intercorrelation matrix produced by these data is characterized by strong, positive correlations in adjacent time periods. As the number of intervening time periods increased, however, the observed correlations systematically decreased and ultimately became negative. Significant negative correlations of earlier performance with later performance are almost never observed with typical time-related performance data. Possible explanations and boundary conditions for these atypical results are discussed.