This paper explores the out-of-sample performances of nine commonly used trading systems with parameter-changing algorithm in six major foreign exchange futures markets. Our research sample covers the period from May 2001 till October 2010 on a weekly basis. In order to correct for data snooping biases, we conduct a stationary bootstrap test developed by White (2000). Considering the proper transaction costs including commissions and slippage, the empirical results indicate that the out-of-sample performances of the best trading systems exhibit statistically significant profitability in most of the foreign exchange futures markets.