Candidate Predictive Features for Crypto Statistical Arbitrage
Summary
This article brainstorms possible inputs for a crypto statistical arbitrage model. It covers relative price moves between similar assets, short and long horizon trends, crowded spreads that may unwind with momentum, lead-lag effects across markets, and seasonal patterns. It also proposes comparing aggressive trading volume with subsequent price changes to infer hidden supply or demand: unusually muted movement despite heavy buying could indicate replenishing sell orders, while the author recalls analysis in which larger-volume moves were more likely to continue.
The article discusses futures carry as another candidate feature, including how a futures premium, funding, and changes in the relationship between spot and perpetual prices may relate to total returns. These are hypotheses and examples, not a tested or combined strategy. One Bitcoin and S&P 500 lead-lag observation was reportedly visible in data examined in 2022, but the author is unsure whether it persists. The piece recommends further analysis of feature strength, decay, and risk-return trade-offs before using these ideas in a model.
Key ideas
- Relative price deviations among similar assets can serve as candidate statistical arbitrage signals.
- Trend, momentum, and lead-lag effects may provide predictive information across different horizons.
- Comparing aggressive trading volume with price response may help identify hidden supply or demand.
- Futures carry and changes in the spot-perpetual relationship are potential model features.
- The proposed features require empirical testing, including analysis of signal strength and decay.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.