We design and deploy a trading strategy that mirrors the Exchange Traded Fund (ETF) arbitrage technique for sector trading. Artificial Neural Networks (ANNs) are used to capture pricing relationships within a sector using intra-day trade data. The fair price of a target security is learnt by the ANN. Significant deviations of the true price from the computed price (ANN predicted price) are exploited. To facilitate arbitrage, output function of the trained ANN is locally linearly approximated. The strategy has been backtested on intra-day data from September 2005. Results are very promising, with a high percentage of profitable trades. With low average trade durations and ease of computation, this strategy is well suited for algorithmic trading systems.