dual momentum backtest

What the numbers mean

This strategy spends most of its life holding one thing and doing nothing. Across a decade of daily bars it makes a handful of decisions, not hundreds, so the record is a few long holds rather than the law of large numbers. The win rate sits near a coin flip, but the sizes are wildly asymmetric: with no stop, no profit target and no holding limit, a winner is kept until its six-month return turns negative, which for a trending asset can take years, while a fader is cut once that same measure rolls over.

The six-month lookback is slow by design: it buys well after a move starts and sells well after it ends, so missing tops and bottoms is arithmetic, not misfortune. The absolute-momentum valve, stand aside when nothing is rising, therefore helps in grinding bear markets and barely at all in sudden ones, since a crash measured in weeks cannot turn a six-month return negative until months later. One position at full notional from a menu of six ETFs makes concentration total, and because this version only reconsiders the menu after an exit rather than monthly, it can ride one holding for years.

Beside this text, read the drawdown as the cost of that slowness, and read the profit factor against the trade count: with this few closed trades, one or two holds carry it. Then ask the harder question, whether the detours into gold, bonds and cash paid for the lag, or whether the window simply rewarded owning stocks.