We review some statistical arbitrage strategies applied to stock markets and investigate their use for trading in commodity markets. A general methodology to develop a statistical arbitrage trading strategy is devised. Cointegration techniques are used to find a long run relationship among commodities and build a portfolio whose value is represented by the deviation from this long run equilibrium. We then verify that the portfolio dynamics contain some predictable components and test trading rules that rely on these predictability properties of the portfolio process. Finally, several profit indicators are proposed in order to measure the performance of the strategy.