Deep Momentum Networks with Market Trend Dynamics

Jaegi Jeon

Abstract:

Time-series momentum (TSMOM) trading strategies manage positions based on the per-
sistence of return trends. Although long short-term memory (LSTM) deep neural archi-
tectures can enhance TSMOM, their performance often deteriorates during abrupt mar-
ket trend changes. This study aims to improve TSMOM performance, particularly at criti-
cal moments marked by significant shifts in long- and short-term trends. To achieve this,
we combine short- and long-term signals into a comprehensive market-state represen-
tation, employing supervised learning to incorporate these market dynamics into the
proposed model. In our experiments, we generate market-state features, referred to as
MTDP scores, by numerically capturing changes in market trends via an extreme gradi-
ent boosting (XGBoost) process. These MTDP scores are then applied within an LSTM-
based trading strategy. A backtest on 99 continuous futures (1995–2021) demonstrates
that incorporating MTDP scores enhances the Sharpe ratio, indicating the effectiveness
of embedding market-state information in TSMOM. Notably, an 8-week fast momentum
look-back window performed best over stable periods (1995–2019). However, during
extreme market downturns, such as the COVID-19 crisis, a 20-week fast momentum win-
dow not only outperformed shorter windows (4- and 8-week signals) but also recovered
more rapidly. These findings suggest that TSMOM strategies can benefit from dynami-
cally adjusting fast momentum windows, consistently generating profitable opportunities
even amid turbulent conditions.