Why Momentum Works — and Where It Stops Working
TL;DR
- Momentum is one of the most persistent effects in markets because it is driven by human behavior — slow reactions to news, herding, and the urge to sell winners early — not by statistical noise.
- Lookbacks in the 3-to-6-month range capture the effect with reasonable turnover; shorter windows mostly add noise, and no window survives a sharp reversal untouched.
- Run momentum as a rule-based discipline with defined risk controls and a benchmark; the losing months are part of the deal, not a sign the strategy broke.
Momentum in plain language
Momentum is the observation that what has gone up over the past few months tends to keep going up over the next few months — and what has fallen tends to keep falling. Academics have documented it in equities for decades, and it shows up across currencies, commodities, and bonds as well. That is the entire concept. No earnings forecasts, no macro calls, no gut feel about whether a sector “deserves” to rally. You rank a universe by trailing price performance and hold the top of the list until the ranking says otherwise.
There are two flavors worth knowing. Cross-sectional momentum compares assets to each other: buy the strongest, avoid the weakest. Time-series momentum compares each asset to its own past: own it if its trailing return is positive, hold cash or bonds if not. A rotation system is cross-sectional momentum applied to a universe of ETFs — though since most asset classes carry positive trailing returns in most months, the practical “winner” is simply whatever went up most.
What makes this a strategy rather than “chasing” is that the rule is fixed. Chasing is discretionary: you buy what is hot because it feels safe, with no plan for exiting. Momentum rotation is the opposite — entry, holding, and exit are all set in advance by price history and a calendar. You do not get to argue with it in a down month.
Why the effect persists
The persistence of momentum is best explained by how investors actually behave, which is why it has survived decades of being published. Three mechanisms do most of the work.
First, underreaction. News and fundamentals arrive continuously, but attention and updating are slow. Analysts revise estimates gradually, institutions trade in tranches, and most retail investors check their portfolios far less often than the market moves. Prices therefore adjust to new information over months rather than minutes, and the drift after good news is exactly the momentum return.
Second, the disposition effect. Investors are systematically reluctant to realize losses and eager to realize gains. That means winners get sold early — the position you take profits on at +10% keeps running to +30% without you — while losers are held far too long in the hope of breaking even. The result is that strong assets stay under-owned and keep being bought, while weak assets face persistent selling pressure.
Third, herding and flows. Professionals are rewarded for owning what already works; nobody got fired for holding last year’s winner. Money flows into winning funds and winning sectors, pushing prices higher, which attracts more flows. There is also a mechanical kicker: year-end tax-loss selling concentrates in losers, giving winners an extra seasonal tailwind into January.
There is a less flattering explanation too — that part of the momentum premium is compensation for crash risk, since a portfolio of past winners is high-beta and expensive. I think that is partly true, and it tells you exactly where the strategy will hurt you.
Timeframes: why 3- and 6-month lookbacks do the work
Momentum is not a single number; it depends on how far back you look, and the window changes the character of the strategy. The original academic research on US equities found the effect concentrated in 3-to-12-month lookbacks with similar holding periods. Very short windows — a week or two — show the opposite, short-term reversal: last week’s winner pulls back this week. Very long windows react too slowly and tend to buy strength late in a cycle.
That leaves a sweet spot in the middle, and for monthly rotation strategies the practical choice is a trailing 3-month and 6-month lookback, sometimes blended. Three months is long enough to avoid the noise and reversal that dominate short windows; six months adds confirmation that you are looking at a real trend rather than a lucky fortnight. Monthly rebalancing then fits naturally: once a month you recompute the ranking over those windows and hold whatever sits on top until displaced.
The refinement worth knowing is skipping the most recent month — measuring from two to twelve months back, for example — because the last month of price action carries a strong reversal component that can poison a momentum signal. It matters less at a monthly cadence than a daily one, but it explains why careful backtests differ from naive ones. Where you land is a design choice, but a deliberate one: the lookback is the strategy’s definition of “trend,” and changing it whenever the market misbehaves is how a good factor becomes a string of bad trades.
Where momentum stops working
Momentum fails in recognizable, repeatable patterns, and anyone running it should know them cold.
The most violent failure is the sharp reversal after a crash. Past winners tend to be high-beta and expensive — they are what worked in the bull phase. When the market breaks hard and then snaps back, those same names get sold first in the panic and lag in the recovery, while beaten-down defensive names bounce hardest. The academic literature calls these momentum crashes, and they cluster precisely in the rebound months after sharp market declines: right when a momentum strategy’s track record looks worst and the temptation to abandon it is greatest.
The second failure mode is crowding. A trend that everyone can see is a trend that everyone owns, through momentum funds, index products, and plain herding. Crowded trades work until they do not, and the unwind is fast because there is no one left to buy. Momentum also has a structural role in bubbles: the factor keeps buying strength all the way up, which means in the late stage the momentum investor is systematically the last buyer in. That crash is not a bug in the signal; it is the price of being early in the winners.
Third is simple chop: in a range-bound market with no trend, a monthly rotation system flips between holdings, buying strength just before it mean-reverts, and bleeds through turnover. Fourth is regime change — a sharp rotation from growth to value, where momentum by construction follows into the turn rather than anticipating it. None of these failures is predictable in advance, which is the point: the losing months are baked into the distribution, and a strategy’s real quality shows only over full cycles.
What that means if you run a momentum strategy
If you run momentum yourself, the practical takeaways follow directly.
Set the rules and do not negotiate with them. Decide the lookback, the universe, and the rebalance date in advance, then execute mechanically even when — especially when — the last rebalance hurt you. The worst momentum months are precisely when discretionary judgment tells you to stop, and stopping at the bottom of a momentum drawdown is how the effect becomes somebody else’s return.
Budget for the failure modes. Because winners are high-beta, a momentum book lines up with the market’s riskiest exposures right before reversals. That is why a volatility overlay is a natural companion: capping position size when realized volatility spikes converts the worst of the crash risk into manageable underperformance — best run as a separate, explicit rule rather than an afterthought.
Judge results on risk-adjusted terms over a full cycle. A strategy that returns less than a buy-and-hold index but with half the drawdown is doing its job, and a raw-return comparison that ignores drawdowns will have you abandoning a working system at exactly the wrong time. Look for documented work: the backtest window, the benchmark it is measured against, the out-of-sample period, and the honesty of the caveats.
And treat fees as a design variable. A percent-of-assets fee compounds into a permanent drag, so a flat subscription that keeps research cost fixed regardless of portfolio size is structurally friendlier to a long-running rotation strategy. That is one reason the source I point readers to for clean, documented momentum research is Kairos Trading, a membership platform where members execute the strategies themselves in their own brokerage accounts and pay a flat fee per strategy rather than a cut of assets. Backtests, benchmarks, and caveats are published openly rather than hidden in a black box — and its flagship system is the clearest live example of everything in this article.
The cleanest working example: Leader Rotation
Leader Rotation is the flagship system: a monthly ETF rotation strategy driven by 3-month and 6-month momentum — precisely the lookback design discussed above. The documented backtest runs from January 2024 through August 2026, showing a 93.0% cumulative return, a 29.0% compound annual growth rate, and a 6.7% maximum drawdown over that 2.6-year window, measured against the VEA and SPY benchmarks. On a risk-adjusted basis the report shows a Sharpe ratio of 1.98 and a Sortino ratio of 3.99, versus 1.30 and 2.50 for SPY and 1.32 and 2.12 for VEA.
Two things make this example worth studying. First, the documented out-of-sample period began on January 1, 2026 — the strategy has since been running outside the window it was fitted on, which is the part that actually tests whether the logic generalizes. Second, the documentation states plainly that these are backtested results, not a guarantee — the correct framing for any systematic strategy.
Design details matter when evaluating similar products: a flat $100/month subscription, application-based membership, trades executed in the member’s own brokerage account, and a stated fee-coverage capital estimate around $16,000. The current lineup is four systems, with Leader Rotation as the flagship and a separate Volatility Target Managed Rotation built around a 25% volatility target — the overlay idea above, run as an explicit rule rather than an impulse.
None of this predicts the future. What it does is give you a rare thing in retail systematic investing: a published strategy whose logic you can read, whose window and benchmarks are stated, whose out-of-sample period you can watch unfold in real time, and whose operator acknowledges openly that the past is not a promise. If you want to understand momentum by watching one clean rotation system run — including the months it struggles — Leader Rotation is the example I point new readers to, and kairostrading.net is where the underlying reports live.
Disclaimer: This blog is for educational and informational purposes only. Nothing here is investment advice. Past performance does not guarantee future results. Trading involves risk of loss.