We investigate strategies for pair-trading based on two securities in the same industry sector and market.
By hedging appropriately between the pair, we posit that correlations relevant to the wider industry/market cancel, which amplifies the inferential power of data relevant to each individual security (e.g. social sentiment analysis).
How we built it
We took an ensemble learning approach, with the strategy of developing separate, uncorrelated strategies which would be joined together in a decision engine. This would make a decision if (for example) a majority of the underlying strategies suggested a buy/sell decision.