Skip to content

Quant Investing: Hidden Markov Model on Winner Portfolios

Having hit its last new high since coronavirus outbreak in 2020, the equity market now is dominated by a few companies, especially in the technology sector, that have driven the rebound. Consequently, technology sector has been the top holdings of momentum ETFs, buying stocks with relatively strong price performance (winner), and results in the strong outperformance of momentum ETFs compared with the market since this April.

Despite the characteristic of momentum that provides an extra boost while markets fall, long-only portfolios may still experience large drawdowns inevitably because of systematic risk. It’s in such high volatile period that we need a robust timing strategy to protect our wealth from systematic risk.

The major contribution of this research are listed as following:

  1. proving that the existence of convexity in the winner portfolios which can be accurately detected by Daniel, Jagannathan, and Kim’s model;
  2. making their model more generalized in the daily data with an extra measure of instant price drift;
  3. the implementation of a Hidden Markov Model (HMM) timing strategy on tradable momentum ETFs with top 3 largest net asset value (NAV): MTUM, PDP and XMMO.
An 8-element parameter vector for HMM.

We proved that our extended model can beat the market empirically with participation in an upward-moving stock market while also offering significant downside protection. The Sharpe ratio of MTUM after timing has increased 43% with near 70% lower max drawdown, and timing PDP’s Sharpe ratio and max drawdown has improved dramatically, 58% higher and 54% lower respectively.

Backtest Result on PDP, a momentum ETF.

A Hidden Markov Model of Momentum, K. Daniel, R. Jagannathan, and S. Kim, Working Paper, 2020

YC Lin
AVP Quantitative Finance, Gamma Paradigm

Contact us at if you want to know more.

Leave a Reply

%d bloggers like this: