Trend following (TF) is technical analysis method for making trading decisions in stock markets. In TF, trend followers buy when the stock price trend raises, and sell when the price trend decreases without predicting the exact price levels. In this paper, we propose an algorithm called trend following with contrarian strategy (TFCS) which combined contrarian strategy and multi-thresholds analysis to determine the buying and selling signals in stock market data. The optimal values of the parameters in TFCS are calculated by the firefly algorithm. The experimental results show that the proposed algorithm can significantly improve the success rate and accumulated return on investment when it is compared to multi-threshold TF.