Momentum Stock: Buying What Already Rose
A momentum stock is a share selected because its recent price performance was strong relative to others. The rule is about past price alone and says nothing about the business, which is why the list changes with the lookback and why the losses arrive together.
How it works
A momentum stock is a share selected because its recent price performance was strong relative to others. The rule looks only at past price; nothing in it examines revenue, margins or debt.
The tendency being exploited is real and reasonably well documented. Relative performance over intermediate horizons has tended to persist, which is why factor investing treats momentum as a durable effect rather than a curiosity.
It is a ranking rule, not a judgement about value. You sort a universe by past return and hold the top slice to the next rebalance. These names are also called relative strength leaders.
The lookback is a free parameter and it decides the list. Rank on the past year and you hold one set of names; rank on the past quarter and you hold another.
The standard construction also skips the most recent month. It ranks on the past several months but excludes the latest stretch, because very short-horizon returns tend to reverse. Almost nobody buying strength applies the skip.
The crash, the crowd and the turnover
It works for years and then fails all at once. The returns are not evenly distributed through time; long favourable stretches are interrupted by sharp reversals that give back a great deal quickly.
The reversal usually arrives when a falling market turns. What fell hardest rebounds hardest, and the ranking is by then holding the other side. The drawdown is sudden rather than gradual.
A ranking rule has to be refreshed to remain one. Names leave the top slice and others enter, so the portfolio turns over on a schedule rather than when a position is finished.
And everyone holds the same names at the same time. The rule is public and mechanical, so it raises the correlation between holdings that look diversified and makes the unwind faster.
Turnover is the cost that decides the result. Each rebalance pays a spread and a commission on every swap, and at small edges that bill can consume the effect entirely.
In practice
Participation confirms it and also crowds it. A leading name usually leads on volume too — the sign the ranking found something real, and the sign everyone else found it.
The effect is measured over months, not days. The documented tendency operates on intermediate horizons, so a daily chart of the same share answers a different question.
And the reversal arrives as a gap, not a drift. Much of the damage lands overnight as an opening gap, through the level you meant to leave at.
So the exit rule matters more than the entry. The entry is only a ranking; the loss comes from holding through the reversal. A trailing stop or a fixed stop loss is the part that decides it.
A trailing exit is a trade-off you can price. On this site’s shared 576-bar history the median position survived 3 bars trailed at one 14-bar average true range and 32 at four, so set risk per trade from the distance you actually choose.
Every round trip costs 2% of a bar. On the same history that is 0.0098 price units, and 45% of the smallest bar.
Specifying it before you start
Fix the lookback and the rebalance interval before you rank anything. Write down how many months of performance you are measuring and the dates on which you will re-sort. Both are the strategy, not settings to adjust once you have seen the results.
Apply the skip. Exclude the most recent month, because that stretch tends to reverse and including it tilts the list toward the names most likely to give it back.
Then count the expected annual turnover cost before deciding whether any of this is viable at your cost level. Estimate how many positions you replace at each rebalance, multiply by the round trip you actually pay, and annualise it. If that bill is a meaningful fraction of the edge you hope to capture, the honest conclusion is that the approach does not work at your size — and arithmetic is a far cheaper way to learn that than a year of trading.
What a momentum stock is not
- Not a growth stock. One is picked on past price, the other on expected earnings.
- Not a verdict on the business. The ranking never opens a financial statement.
- Not trend following. That asks whether a market trends; this asks which has risen most.
- Not a permanent property of a share. A name is a momentum name until the next re-ranking.
When it fails
In a range it buys high and sells low repeatedly. A ranking rewards whatever has moved most, and inside a trading range that is simply whatever sits nearest the top of it.
- When the movement has no follow-through. The ten-bar efficiency ratio on this history has a median of 0.34, with 30% of bars above 0.5 — most travel is round trip rather than progress.
- When the breakout that qualified the name does not hold. A close above a 20-bar high closed back below it within ten bars in 85% of 39 events, and above a 55-bar high in 100% of 11.
- When the market turns. The losses arrive together, in the names everyone held, so a spread of top-ranked positions is not the diversification it appears to be.
- When the lookback was chosen after seeing the results. Testing several windows and keeping the best is overfitting, and a ranking rule is unusually easy to fit.
- When the skip is dropped. Ranking on the most recent stretch loads the list with whatever has just run, which is the part most likely to reverse.
The original data
Seven videos carry “momentum stock” in the title, at a median of 908 views and a maximum of
1,850,711. The scan of 31,760 videos in research/search-study-corpus.jsonl, logged in
research/broker-coverage.json, spreads those seven across 6 channels — the widest gap between
median and maximum in this part of the corpus. One found an enormous audience and the rest found
almost none, which is what a category looks like when it is watched for a story rather than a method.
Then the run lengths, which say why the horizon matters. research/series-measurements.json, via
site/measure_series.py, puts direction runs on this site’s shared 576-bar history at an average of
2.01 bars with a longest of 11 across 286 runs, so short-horizon continuation is the exception. The
measured effect is a monthly one, and applying it to a daily chart is applying it to a different
question. Write down the lookback, the skip and the rebalance date before you rank anything; choosing
them after seeing the list is how a ranking rule becomes a story.
Related
Momentum trading is the procedure these shares sit inside — the ranking, the rebalance and the rotation that turn a list into a portfolio.
Factor investing treats momentum as one effect among several, and it is where the skip and the rebalance conventions come from.
Trend following asks a different question of the same chart — whether a market is trending at all, rather than which name has risen most.
Buying something at a new high has never stopped feeling wrong to me. Every instinct I have says wait for a pullback, and the pullback is exactly what the rule tells me not to wait for. What helped was noticing that the discomfort is not information - it is the chart arriving after the move instead of before it. I still feel it. I have stopped treating it as a reason.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.