Knowledge · Strategies · Dual Momentum (GEM)

Dual Momentum (GEM)

Gary Antonacci - 'Dual Momentum Investing' (2014)
Momentum Evidence: Strong position equitiesglobal_indices
5/10
Relevance for Botty
Combines absolute + relative momentum: long the asset with the highest 12-month return, but only if that return > cash/bonds. Otherwise bonds.
Core concept
Global equities are ranked monthly by 12-month performance. Long the best one. Additional filter: only if the best one is also absolutely positive versus bonds - otherwise fully into bonds. A combination of relative + absolute (time-series) momentum.
Relevance for Botty
Relevance Score 5/10
Botty is single-asset BTC perp intraday - Dual Momentum on a monthly basis does not fit directly. But: the concept of 'only trade when the macro trend is also absolutely positive' is transferable to BTC as an optional long-only filter (e.g. 30-day return > 0).
Rules

Entry

  • At the end of each month: compute the 12-month return for each candidate (e.g. S&P 500, MSCI EAFE, bonds)
  • Pick the equity asset with the highest return (relative momentum)
  • Check: is its 12M return > the T-bill return? (absolute momentum)
  • If yes -> 100% into the equity asset; if no -> 100% into aggregate bonds

Exit

  • Re-evaluate at month-end - on a signal change, reallocate completely
Parameter
NameTyp. valueDescription
lookback_months 12 Window for the momentum ranking
rebalance monthly Frequency
universe_size 2-3 equity indices + 1 bond Antonacci uses S&P 500, MSCI EAFE, Barclays Agg
Pros & Cons

Pros

  • Very low trade frequency (2-3 trades/year)
  • Clear binary signals, simple to code
  • Drawdown substantially reduced vs. buy-and-hold
  • Confirmed by a 215-year backtest

Cons

  • Underperforms in V-shaped rebound rallies (re-enters too late)
  • Endogenous risk: after a large drawdown it is stuck in 'safe' bonds and misses the bounce
  • Operates at monthly granularity; misses intra-month moves
Typical performance
cagr
~15% 1974-2013 vs. ~10% S&P 500
notes
A 215-year backtest confirms the effect. Outperformed in bear markets (1973-74: +20% vs. -40% S&P).
sharpe
~0.8-1.0 historically
max drawdown
~20% vs. ~50% S&P 500
trades per year
2-3
Bot suitability
Ideal for low-frequency position-trading bots. No deep infrastructure required.
Background

Core idea

Dual Momentum combines two classic effects:

  1. Relative momentum (Jegadeesh/Titman 1993): past winners beat past losers.
  2. Absolute momentum (time-series momentum): the asset must additionally have performed positively versus cash/bonds.

The idea: both filters together pull you out of bear markets in particular, because after deep drawdowns even the 'relative winner' is often still absolutely negative - then Dual Momentum moves into bonds and waits.

The GEM rule (Global Equities Momentum)

At month-end:

  1. Compute the 12-month return for S&P 500, MSCI EAFE (non-US equities) and aggregate bonds.
  2. Which equity asset has the higher 12M return? -> candidate.
  3. Is the candidate's 12M return > T-bills?
  4. Yes: 100% into the candidate equity.
  5. No: 100% into aggregate bonds.

The result: 2-3 trades per year, long holding phases, substantially reduced drawdown.

Why it works

The absolute-momentum filter is the key. In 2008 the relative winner was still strongly negative -> into bonds, missing the crash. In 2020 it flips quickly back into equities as soon as the macro trend turns. In the 1970s bear markets (long, grinding) it sat in bonds until the absolute filter turned.

Criticism

  • V-shaped recoveries like March 2020 are captured poorly - by the time the 12M return turns positive again, a 30% rally is over.
  • The monthly cadence is arbitrary; daily/weekly variants perform similarly or better.
  • On pure US-equity history, GEM is barely superior to a simple 200-MA filter.

Relevance for Botty

Botty is a single-asset intraday bot - Dual Momentum in the original GEM format does not fit. The concept of the absolute filter is, however, valuable: only allow long trades when BTC's 30-day return is positive. This is a very simple rule that historically filters out long signals in downtrend regimes - exactly what Dual Momentum does at a coarser granularity.