Pairs Trading, Technical Analysis and Data Snooping: Mean Reversion vs Momentum

Traders typically lock in gains and gamble on losses, as prospect theory suggests. The disarmingly simple idea behind spread trading can lead to an informationless investing strategy with potentially delusive performance and a biased estimation of risk. In this paper, we develop a two-layer approach for evaluating the out-of-sample performance of technical analysis on frequently traded spreads using daily data from 1990 to 2016, while testing for presence of manipulative concave payoffs. The empirical evidence suggests that any short-term outperformance is mainly manipulation-free and rejects the existence of a uniformly monotonic downward trend over the years. For commodity spreads, the evidence of significant and skilled returns appears stronger.