This article summarises published academic research and our own hypothetical backtest results. Neither is a guarantee of any future outcome, and nothing here is investment advice.
Momentum is the best-documented anomaly in equities, and it owns the ugliest tail in the factor literature. The classic winners-minus-losers portfolio lost roughly 91% in two months of 1932 and roughly 73% across three months of 2009. Those episodes have a name, a mechanism, and a degree of predictability, and any momentum system that has not planned for them is a system that has not read its own literature.
What a momentum crash is
The academic momentum factor is a long-short construction: buy the stocks with the strongest trailing returns, short the weakest, rebalance monthly. Daniel and Moskowitz’s 2016 study of that portfolio found that its worst months are not scattered randomly through history. They cluster in one specific environment: the violent rebound that follows a bear market, arriving when trailing volatility is high and the market rips upward off a distressed low.
The engine of the crash is the short leg. After a long decline, the “loser” portfolio is full of beaten-down, high-beta, often near-bankrupt names. When the panic reverses, exactly those names rally hardest, sometimes doubling in weeks. The momentum trader is short that rally. Daniel and Moskowitz describe the position as carrying embedded optionality: in a rebound, the short side behaves like a written call option on the market, and the strategy’s losses compound just as the broad tape turns euphoric. March to May 2009 is the canonical modern case; July and August 1932 is the historical one.
The crash is partly forecastable, which is the interesting part
Two robust findings follow. First, crash risk is state-dependent: it concentrates after bear markets and under high volatility, so a simple pair of conditioning variables (a bear-market flag and trailing volatility) identifies most of the danger zone in advance. Second, acting on that state helps. Barroso and Santa-Clara (2015) showed that momentum’s own volatility is highly persistent and therefore forecastable, and that scaling the position down when forecast volatility is high roughly doubled the strategy’s Sharpe ratio in their sample while taming the worst episodes. Daniel and Moskowitz reach a similar place with a dynamic weighting scheme. The lesson generalises: the momentum premium does not pay you for holding it blindly through every regime, and the regimes are partially observable.
What changes when you are long-only
A long-only momentum book, which is what our system runs, cannot be squeezed on a short leg, so the textbook crash mechanism does not apply one-for-one. It would be convenient to stop there. It would also be wrong. The long-only analogue of a momentum crash is relative and absolute pain at the same regime turn: the recent winners you hold stall or fall while yesterday’s junk doubles, and the factor tailwind you have been compounding on inverts just as volatility peaks. The drawdown is smaller than the long-short disaster, but it lands at the same calendar spot, and a leveraged or concentrated long book can still be ruined by it.
How our system addresses the same risk
Shishin’s answer is structural rather than heroic. A breadth-driven regime gate decides each day whether the environment supports small-cap momentum at all, and deploys the momentum engine only when breadth is broad and rising. When conditions turn hostile the system rotates to defensive engines or stands down entirely; it held cash on roughly a third of all days in the five-year backtest window. That is the same design intuition as the academic vol-scaling result, implemented as a deterministic rule rather than a fitted model (we tested learned regime models against the gate and published why they lost). Over the locked window the full stack’s worst drawdown was −15.9%, hypothetical and documented trade by trade.
None of that is immunity. The gate reacts to conditions it can observe, which means it is late by construction, and a fast enough regime turn will always land the first blow. The honest claim is narrower: momentum crash risk is real, it is state-dependent, and a system that conditions its exposure on the state has the literature on its side.
Sources & further reading
- Daniel, K. & Moskowitz, T. J. (2016). “Momentum Crashes.” Journal of Financial Economics, 122(2), 221 to 247.
- Barroso, P. & Santa-Clara, P. (2015). “Momentum Has Its Moments.” Journal of Financial Economics, 116(1), 111 to 120.
- Jegadeesh, N. & Titman, S. (1993). “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency.” Journal of Finance, 48(1), 65 to 91.