Slow Momentum with Fast Reversion: A Trading Strategy Using Deep Learning and Changepoint Detection

Momentum strategies are an important part of alternative investments and at the heart work commodity trading advisors. These have, however, been found to have difficulties adjusting rapid changes in market conditions, such as during 2020 crash. In particular, immediately after momentum turning points, when a trend reverses from uptrend (downtrend) downtrend (uptrend), time-series prone making bad bets. To improve responsiveness regime change, authors introduce novel approach, which they insert online changepoint detection (CPD) module into deep network pipeline, uses long short-term memory deep-learning architecture simultaneously learn both estimation position sizing. Furthermore, their model is able optimize way it balances (1) slow strategy that exploits persisting trends but does not overreact localized price moves (2) fast mean-reversion by quickly flipping its then swapping back again exploit moves. The CPD outputs location severity score, allowing respond varying degrees disequilibrium, or smaller more changepoints, data-driven manner. test over period 1995–2020, addition leads 33% improvement Sharpe ratio. especially beneficial periods significant nonstationarity; most recent years tested (2015–2020), performance boost approximately 66%. This interesting because traditional underperformed this period.