Triangulated Statistical Arbitrage from Overlapping Pairs
Summary
This installment proposes converting signals from overlapping pairs into security-level signals. For each spread, its z-score becomes two opposing votes: the relatively rich ticker receives a positive signal and the relatively cheap ticker a negative one. Averaging the votes across each ticker can reinforce a mispricing supported by several partners while weakening a divergence that appears isolated. The example uses energy stocks and describes a network in which XOM is stretched against several peers while the other spreads are near their usual levels.
The resulting ticker scores can guide a long/short portfolio, with cheap names in the long book and rich names in the short book. Portfolio construction can target roughly neutral market exposure or allow net exposure to vary with the relative signals. The proposed benefits include using more pair research, directing capital at the suspected mispriced legs, and retaining some variance control through portfolio hedging. These are conceptual claims and an illustrative example, not reported backtest results. The article also flags unstable betas, measurement error, noisy moves, and sparse network connections; it defers methods for judging signal agreement to a later installment.
Key ideas
- Each spread z-score can be converted into equal and opposite signals for its two securities.
- Aggregating signals across overlapping spreads can strengthen consensus and reduce the influence of conflicting votes.
- Ticker-level signals can form a long/short portfolio of securities judged cheap or rich across the network.
- Portfolio weighting can control net market exposure or allow exposure to vary with signal strength.
- Unstable betas, measurement error, noise, and sparse connections can make the aggregated signal less reliable.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.