Skip to content
Menu
Finabase Learn

Strategies · Trend and momentum

52-week high momentum: buying what is near its high, tested

The 52-week high strategy buys what trades closest to its one-year high. What the research found, and a 27-year test on sector funds where it lost to holding them all.

On this page
  1. The idea
  2. What can and cannot be tested here
  3. The rules
  4. What the backtest shows
  5. Where it fails: the losing periods
  6. Compared with its neighbours
  7. How Finabase does it
  8. Frequently asked questions
  9. Summary
  10. Check your understanding

Every stock quote page prints a 52-week high, and most people read a price near it as "expensive". In 2004 two finance professors, Thomas George and Chuan-Yang Hwang, reported the opposite. Across US stocks from 1963 to 2001, the stocks trading closest to their 52-week highs went on to do better over the next six months than the stocks trading farthest below them, and this simple measure predicted returns better than the past return that ordinary momentum strategies rank on.

This lesson explains the idea, states honestly that the published strategy cannot be reproduced with the data behind these pages, and tests the nearest version that can: ranking nine sector funds by how close each is to its one-year high. In that test the rule lost to simply holding all nine. The reverse rule, buying the funds farthest from their highs, did about as well as the market. The page reports both.

The idea

Ordinary relative momentum ranks stocks by how much they have risen over the past six or twelve months. The 52-week high version ranks them by a different number:

nearness = current price ÷ highest price of the past 52 weeks

A stock at $95 with a one-year high of $100 has a nearness of 95%. A stock at $60 with the same high has 60%. The measure ignores where the price was a year ago. A stock can be up 40% on the year and 20% below its high, or flat on the year and at its high.

George and Hwang's explanation is anchoring. The 52-week high is published everywhere, and traders use it as a reference point. When good news pushes a stock toward its high, some holders sell because the price "is at its high", and buyers hesitate for the same reason. The price adjusts to the news too slowly. Once it breaks through, the rest of the adjustment follows. If that is right, being near the high marks a stock whose good news is not yet fully in the price.

Their paper reports that a portfolio buying the 30% of stocks nearest their highs and selling short the 30% farthest away earned more than the standard momentum portfolios of Jegadeesh and Titman over 1963–2001, and that its gains did not reverse in later years.

Close as a share of the 12-month high: technology (XLK), energy (XLE) and utilities (XLU), 1999–2026 50% 60% 70% 80% 90% 100% XLK XLE XLU 2000 2005 2010 2015 2020 2025
Month-end close as a share of the highest month-end close of the past 12 months, for three of the nine sector funds. A line at 100% means the fund closed the month at a one-year high.

What can and cannot be tested here

The published strategy needs three things these pages do not have:

  • Every listed stock, including those that later vanished. The data here covers funds and current large companies. A test on today's survivors flatters any stock-picking rule.
  • Short selling. Part of the published return comes from the short side, stocks far below their highs, with borrowing costs the paper does not charge.
  • Daily highs over 52 weeks for decades. The long histories here are monthly closes.

So this page tests two substitutes and labels them as such. The main one uses nine sector funds with monthly closes back to 1998, long only. It answers a narrower question: among nine diversified funds, did nearness to the one-year high pick the better ones? The second uses 30 individual stocks over four years.

Neither is a test of the paper. A fund of 30 to 80 stocks has much less company-specific news than one stock, and anchoring on a fund's high is a weaker idea than anchoring on a stock's.

The rules

On the last trading day of each month:

  1. For each of the nine sector funds (XLK, XLF, XLE, XLV, XLU, XLP, XLY, XLI, XLB), divide the month-end close by the highest month-end close of the last 12 months, the current month included.
  2. Rank the nine from nearest to farthest. When several are exactly at their high, the one with the larger 11-month return ranks first.
  3. Hold the top three in equal parts. Sell any fund that has left the top three and buy its replacement at the same close.
  4. Do nothing until the next month-end.

The rule is always fully invested. It never goes to cash, so it is a selection rule and not a market-timing rule.

A worked example

At the end of June 2008, energy (XLE) closed the month at its 12-month high: nearness 100%. Utilities (XLU) were 4.7% below theirs and materials (XLB) 6.2% below. Financials (XLF) were 41.0% below. The rule held XLE, XLU and XLB.

It was the wrong moment to hold energy and materials. Commodity prices peaked that summer and both funds fell hard in the second half of the year. The rule lost 31.2% in 2008. Holding all nine lost 37.0%, so the rule did slightly better, but being near a high gave no protection.

At the end of December 2022 the top of the ranking was health care (XLV, 2.3% below its high), then energy and consumer staples (both 4.0% below). Consumer discretionary (XLY) was 30.2% below. In 2023 the market's gains came from technology and consumer discretionary, the funds the ranking had placed last. The rule lost 2.6% that year while all nine gained 11.5%.

What the backtest shows

The test, stated in full:

  • Data. Monthly closes for the nine Select Sector SPDR funds that have traded since December 1998. The first 11 months build the 12-month high, so the test runs from December 1999 to October 2026: 322 months, 26.8 years.
  • Trading. Ranked and traded at the month-end close.
  • Costs. 0.10% of the amount on every buy and every sell.
  • Comparison. All nine funds in equal parts, rebalanced monthly with the same cost. This is the fair benchmark: the same nine funds with no selection.
  • Left out. Dividends and taxes. Sector funds pay different dividends, so leaving them out affects the rule and the benchmark by different amounts; the page does not measure that.
Nine sector funds monthly, 1999–2026: the three nearest their 12-month high against all nine, 0.10% per trade, no dividends $1 $2 $5 Growth of $1 All nine sectors, equal parts Three nearest their high 0% −20% −40% −60% Fall from the peak 2000 2005 2010 2015 2020 2025
Nine sector funds, monthly, December 1999 to October 2026: the three nearest their 12-month high against all nine in equal parts, 0.10% per trade, dividends not included. Top: growth of $1 (log scale). Bottom: the fall from the highest value so far.
Measure Three nearest their high All nine, equal parts
Total return 170.7% 476.8%
Return per year (CAGR) 3.8% 6.7%
Deepest fall from a peak −45.0% −50.6%
Longest time below a peak 85 months 62 months
Portfolio turnover per year 4.1 times 0.2 times
Total paid in costs 30.7% 1.8%

SPY over the same months returned 6.3% a year, with a deepest fall of 52.2%.

The rule lost by 2.9 percentage points a year. Compounded over 27 years, $1 grew to $2.71 under the rule and to $5.77 holding all nine.

Costs explain under a third of the gap. The holdings changed in 235 of 322 months, and the portfolio turned over about four times a year. With no costs the rule returns 4.6% a year, still well short of 6.7%. The selection itself was worse than no selection.

The ranking is unstable. In a rising market several funds sit within a percent or two of their highs, and which three are nearest changes on small differences. Every fund was held between 26% and 41% of the time. A rule that keeps swapping between near-equal candidates pays costs for changes that carry no information.

The two halves of the period

Period Three nearest their high All nine, equal parts
December 1999 – February 2009 −5.5% a year −3.9% a year
February 2009 – October 2026 9.0% a year 12.7% a year

It trailed in both. The split is at the bottom of the 2008 fall, so the first half contains two bear markets and the second a long rise. Neither kind of market favoured the rule.

Where it fails: the losing periods

The years it trailed most.

Year Three nearest their high All nine What happened
2001 −22.0% −8.2% Held the sectors that had led into the peak
2006 1.0% 14.1% Leadership changed several times during the year
2009 11.8% 25.0% The rebound was strongest in funds far below their highs
2023 −2.6% 11.5% Held defensive funds while technology recovered
2025 −2.4% 11.3% Repeated switches around the spring fall

The rule had good years too: 35.5% against 28.1% in 2013, and 19.8% against 7.3% in 2020.

Recoveries. After a deep fall, the funds nearest their highs are the ones that fell least: utilities, staples, health care. The funds that rebound most are the ones that fell most, and they rank last. The rule is built to miss the first leg of every recovery.

The opposite rule did better. Holding the three funds farthest from their highs returned 6.3% a year, close to the market, though with a deepest fall of 62.2%. If nearness to the high carried the information the stock research found, the far group should have done clearly worst. Among sector funds it did not.

How many funds are held.

Funds held Return per year Deepest fall Turnover per year
1 1.0% −54.2% 6.8 times
2 1.9% −47.1% 5.6 times
3 3.8% −45.0% 4.1 times
4 6.1% −44.8% 3.0 times
5 6.7% −42.6% 2.3 times

The pattern is consistent: the more selective the rule, the worse it did. Results improve only as the portfolio approaches "hold everything". That is evidence against the selection, not a reason to choose five.

Individual stocks. The same idea on the 30 stocks in the Dow Jones Industrial Average as of late 2024, using weekly closes from November 2022 to November 2026, a 52-week high, and a new ranking every four weeks:

Portfolio Return per year Deepest fall
Six stocks nearest their high 6.1% −17.9%
Six stocks farthest from it 17.0% −16.5%
All 30 in equal parts 16.3% −13.4%

Again the near group trailed. Four years and 30 survivors is a small, biased sample, and it says nothing about the thousands of smaller stocks where the published effect was measured. It does show that the idea is not something a reader can expect to see working in a handful of large companies.

Other limits:

  • Taxes. Monthly switching makes most gains short-term in a US taxable account (IRS Topic 409).
  • Published effects fade. Momentum effects in US stocks have been weaker since they became widely known; see Why trading strategies stop working.
  • The research result is long-short. A long-only investor holds only the near group and gets only that part of the published return.

Compared with its neighbours

Rule Ranks by Can hold cash Tested here on Against its benchmark
52-week high momentum Price ÷ one-year high No Nine sector funds Trailed
Relative strength rotation Return over recent months No Five index funds See that lesson
Time-series momentum Each fund's own past return above zero Yes One fund at a time See that lesson
Donchian channel breakout A new high of the last so many months Yes One fund at a time See that lesson

The channel breakout also acts on highs, but it asks a yes-or-no question about one symbol (is this a new high?) and holds cash otherwise. The 52-week high strategy compares many symbols with each other and is always invested.

How Finabase does it

No Finabase tool ranks a portfolio by nearness to the high or tells you what to hold.

  • Quote pages show the 52-week high and low for every symbol, and the chart link above shows how far a fund sits below its high over five years.
  • Screener. Stock screener results can show the 52-week high in a column beside the price, which are the two numbers this strategy divides. A screen lists what meets a condition today; it does not test what happened afterward.
  • Backtest and Strategy. The Backtest tool trades indicator signals on one symbol and has no ranking across symbols. The Strategy tool replays if/then allocation rules over two years of daily prices. Backtest buys a fixed $10,000 on each signal and charges no trading cost, so its results are not comparable with the table on this page, which charges 0.10% a trade over 27 years.

Frequently asked questions

Is a stock near its 52-week high expensive?

Nearness to the high says nothing about value. It compares today's price with the past year's prices, not with earnings or assets. The research found that, on average across many stocks, nearness to the high was followed by better returns, not worse. This page's tests on funds and large stocks did not find that.

Why did the test here disagree with the published research?

The tests differ in what they hold. The paper covers thousands of stocks, many of them small, buys the near group and sells the far group short, and ends in 2001. This page holds nine diversified funds, long only, through 2026. The effect may be specific to individual stocks, may have faded, or both. This data cannot separate those.

How is this different from buying breakouts?

A breakout trader buys one stock when it passes a level and uses a stop. The 52-week high strategy is a ranking: it holds whatever is nearest its high relative to the alternatives, whether or not anything has broken out, and has no stop.

Does the 52-week low work the other way?

In the published research the stocks farthest from their highs did worst. In this page's sector test the far group did about as well as the market, with deeper falls. Buying what has fallen most is a different strategy, covered under Short-term reversal: buying last week’s losers, tested and Mean reversion vs trend following: two rules on one market.

Summary

The 52-week high strategy ranks by price divided by the past year's high and holds what is nearest. Research on US stocks from 1963 to 2001 found that this ranking predicted returns better than past returns did. That strategy needs thousands of stocks and short selling and cannot be tested here. The nearest testable version, the three of nine sector funds closest to their 12-month high, returned 3.8% a year from December 1999 to October 2026 after costs, against 6.7% for holding all nine. It trailed in both halves of the period, costs explain under a third of the gap, and the funds farthest from their highs did better. Among sector funds and large stocks, nearness to the high did not select winners.

Check your understanding

5 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. Stock A trades at $90 with a 52-week high of $100. Stock B trades at $48 with a 52-week high of $50. Which ranks higher in a 52-week high strategy?

    1. Stock A, because its price is higher
    2. Stock B, because it is 4% below its high and A is 10% below
    3. They rank equally
    Show the answer

    B. Stock B, because it is 4% below its high and A is 10% below The ranking uses price divided by the 52-week high: 96% for B and 90% for A. The price level and the past year’s return do not enter.

  2. Why can the strategy published by George and Hwang not be tested with the data behind this page?

    1. It needs thousands of stocks including ones that no longer trade, and short selling
    2. The paper’s rules are secret
    3. It only works on monthly bars
    Show the answer

    A. It needs thousands of stocks including ones that no longer trade, and short selling The paper ranks every listed US stock and sells the far group short. The data here covers funds and current large companies, so the page tests sector funds as the nearest version and says so.

  3. In the sector test, the three funds farthest from their highs returned 6.3% a year and the three nearest returned 3.8%. What does that say?

    1. Among these funds, nearness to the high did not pick the better performers
    2. The far group is the better strategy
    3. The test contains an error
    Show the answer

    A. Among these funds, nearness to the high did not pick the better performers If the ranking carried information, the far group should have done clearly worse. It did not, so the ranking failed in this sample. That does not make the far group a strategy either: its deepest fall was 62.2%.

  4. With no trading costs the rule returns 4.6% a year instead of 3.8%, against 6.7% for holding all nine. What follows?

    1. Costs are the whole problem
    2. Costs explain part of the gap; the selection itself also trailed
    3. Costs do not matter for monthly strategies
    Show the answer

    B. Costs explain part of the gap; the selection itself also trailed Removing costs closes 0.8 of the 2.9 percentage points. The rest is the funds chosen. Turnover of four times a year made the cost visible, but it is not the main cause.

  5. After a deep market fall, which funds does the rule hold?

    1. The funds that fell most, because they are cheapest
    2. The funds that fell least, which sit nearest their highs
    3. Cash
    Show the answer

    B. The funds that fell least, which sit nearest their highs Nearness to the high is greatest for whatever fell least. Those funds usually rebound least, which is why the rule trailed in 2009 and 2023. It never holds cash.

Sources and further reading

  1. Thomas J. George and Chuan-Yang Hwang, “The 52-Week High and Momentum Investing,” Journal of Finance 59(5), 2004, pp. 2145–2176
  2. Narasimhan Jegadeesh and Sheridan Titman, “Returns to Buying Winners and Selling Losers,” Journal of Finance 48(1), 1993
  3. IRS — Topic no. 409, Capital gains and losses
  4. Finabase chart data — monthly bars for nine Select Sector SPDR funds, December 1998 to October 2026, and weekly bars for 30 Dow stocks, November 2021 to November 2026; rules and results computed by Finabase