Skip to content
Menu
Finabase Learn

Strategies · Trend and momentum

Time-series momentum: the 12-month return rule, tested

Time-series momentum holds a fund only when its own return over the past year is positive. The exact rule, a 33-year SPY backtest with costs, and where it lost.

On this page
  1. The idea
  2. The rules
  3. What the backtest shows
  4. Where it fails: the losing periods
  5. Compared with its neighbours
  6. How Finabase does it
  7. Frequently asked questions
  8. Summary
  9. Check your understanding

Time-series momentum asks one question about one asset: is it higher than it was a year ago? If yes, hold it. If no, hold cash. The asset is compared only with its own past, which is why the rule is also called absolute momentum. It is different from the better-known kind of momentum that ranks assets against each other and buys the strongest.

Moskowitz, Ooi and Pedersen gave the rule its name in a 2012 paper that tested it on 58 futures markets. This lesson states the single-fund version exactly and tests it on 33 years of SPY, the oldest S&P 500 exchange-traded fund. It behaves almost like the 10-month moving average rule: the two agreed in 349 of 393 months. The result has the same two halves, a large lead through 2009 and a large deficit since.

The idea

Prices that have been rising tend, a little more often than not, to keep rising for some months, and prices that have been falling tend to keep falling. Researchers call this momentum. It has been measured in stocks, bonds, currencies and commodities.

The usual explanation is that people react to news slowly and then, once a trend is obvious, pile in. Both habits stretch a move out over months. Nobody has shown that this must continue, and the effect is weak in any single month. A rule built on it wins by being on the right side of a few long moves, not by being right often.

Time-series momentum is the plainest possible use of the idea. It makes no forecast of how far a move will go. It measures one thing, the sign of the past year's return, and acts on it.

How it differs from a moving average rule

A moving average compares today's price with the average of the last ten months. Time-series momentum compares it with one price, the one from twelve months ago. The two usually agree, because a price above where it stood a year ago is usually above its recent average as well.

They disagree at turns. The 12-month rule depends on a single old price, so its answer can flip because of what happened a year ago rather than what happened this month. If the market fell sharply twelve months back, that low price leaves the comparison and the hurdle jumps.

The rules

Checked once a month, on the last trading day:

  1. Take today's close and the close on the last trading day of the same month one year earlier.
  2. If today's close is higher, hold the fund for the next month.
  3. If it is the same or lower, hold cash for the next month.
  4. Do nothing between month-ends.

There is nothing else. The only choice in the rule is the lookback, 12 months, and the test below shows what other lookbacks did.

SPY month-end close and its 12-month return, 1994–2026: shaded months are in cash $50 $100 $200 $500 SPY month-end close (log scale) -0.5 0 0.5 12-month return 1995 2000 2005 2010 2015 2020 2025
SPY month-end close (top, log scale) and its 12-month return (bottom), 1994 to 2026. The rule holds cash in the shaded months, when the 12-month return is at or below zero.

A worked example

At the end of January 2008 SPY closed at $137.37. Twelve months earlier it had closed at $143.75. The 12-month return was −4.4%, so the rule moved to cash.

It stayed in cash through the rest of the bear market. At the end of October 2009 SPY closed at $103.56, and a year earlier it had closed at $96.83: a return of 7.0%. The rule bought. It had sold at $137.37 and bought back at $103.56, 24.6% lower.

The re-entry shows the "old price" effect. October 2008 was the month of the crash, so one year later the comparison price dropped sharply and the hurdle became easy to clear. The rule bought eight months after the lowest month-end close, in February 2009.

What the backtest shows

The test, stated in full:

  • Data. SPY month-end closes from February 1993 to October 2026. The first 12 months supply the first comparison, so the test runs from February 1994: 393 months.
  • Trading. The signal and the trade both use the month-end close.
  • Costs. 0.10% of the amount on every buy and every sell.
  • Left out. Dividends, interest on cash and taxes. Buy and hold collects every dividend; the rule collects them 78% of the time and would have earned interest on cash the rest. The original paper measured returns above the Treasury bill rate; this test compares with zero.
SPY monthly, 1994–2026: 12-month momentum rule against buy and hold, 0.10% per trade, no dividends $1 $2 $5 $10 Growth of $1 Buy and hold 12-month momentum rule 0% −20% −40% −60% Fall from the peak 1995 2000 2005 2010 2015 2020 2025
SPY monthly, 1994 to 2026: the 12-month momentum rule against buy and hold, 0.10% per trade, dividends not included. Top: growth of $1 (log scale). Bottom: the fall from the highest value so far.
Measure 12-month momentum rule Buy and hold
Total return 1,333.4% 1,514.4%
Return per year (CAGR) 8.5% 8.9%
Deepest fall from a peak −20.8% −52.2%
Longest time below a peak 41 months 80 months
Time invested 78% 100%
Round-trip trades 14 1
Trades with a gain 50% n/a

Half the trades lost money, and the rule still came close to buy and hold. The seven gains averaged 56.7% and the seven losses 2.5%. Two trades did most of the work: February 1995 to November 2000 gained 169.3%, and the position bought in April 2023 was up 81.5% at the end of the test. A strategy with this shape cannot be judged by its win rate.

The two halves of the period

Period 12-month momentum rule Buy and hold
February 1994 – February 2009 8.6% a year 3.1% a year
February 2009 – October 2026 8.4% a year 14.1% a year

From August 2000 to September 2002 the rule lost 13.3% and buy and hold lost 46.3%. From October 2007 to February 2009 the rule lost 11.3% and buy and hold 52.2%. Those two stretches are the whole of the rule's advantage. Since February 2009 it has earned 5.7 points a year less than holding.

Where it fails: the losing periods

The rule loses when the market drops below its level of a year earlier and recovers within a few months. It sells after the fall and buys back after the rebound.

Sold Bought back Result
September 2011, $113.15 October 2011, $125.50 Bought back 10.9% higher
August 2015, $197.67 October 2015, $207.93 Bought back 5.2% higher, sold again in December
December 2018, $249.92 February 2019, $278.68 Bought back 11.5% higher
March 2020, $257.75 May 2020, $304.32 Bought back 18.1% higher
April 2022, $412.00 April 2023, $415.93 Out for a year; back 1.0% higher

The rule's deepest fall, 20.8% from September 2018 to May 2020, came from two of these in a row: it sold in December 2018 near the low, bought back higher, held through the fall of February and March 2020, and sold at the March close.

2022 is the instructive case. The rule sold in April 2022 at $412.00, avoided the rest of that year's decline, and stayed out until April 2023, when it bought at $415.93. It skipped a further fall of 13% to the September 2022 close of $357.18 and gained nothing from doing so, because the market had fully recovered before the 12-month return turned positive.

Other ways it fails or misleads:

  • A fall inside the lookback's blind spot. The rule looks once a month at one comparison. A crash that starts and ends between two month-ends is taken in full.
  • Taxes. Each sale at a gain is taxed in that year in a US taxable account, and positions held a year or less are taxed as ordinary income (IRS Topic 409). Buy and hold defers the tax.
  • It is one rule on one market here. The 2012 paper's results came from spreading the rule over 58 markets, long and short, with each position sized by its volatility. A single stock fund, long or cash, is a much narrower thing.
  • Since the paper. The paper's sample ended in 2009, the best possible date for a trend rule on stocks. The second row of the split table is what came after.

Neighbouring lookbacks

Lookback Return per year Deepest fall Round trips Trades with a gain
3 months 6.2% −32.1% 41 56%
6 months 7.9% −29.8% 24 67%
9 months 8.5% −24.9% 18 61%
12 months 8.5% −20.8% 14 50%

The shorter the lookback, the more trades and the worse the result, on both return and the deepest fall. Nine and twelve months are close. This is consistent with the research, which finds the effect strongest around a year, but four rows on one history are thin evidence. Note that the row with the lowest share of winning trades had the best outcome.

Other markets

The same 12-month rule on six other funds, each from the first month its data allows to October 2026:

Fund From Rule, per year Held, per year Rule's deepest fall Held
QQQ (Nasdaq-100) 2000 9.3% 7.8% −38.2% −79.6%
IWM (small companies) 2001 4.5% 6.9% −31.0% −53.5%
EFA (developed markets) 2002 4.1% 5.0% −36.7% −59.7%
EEM (emerging markets) 2004 5.9% 5.7% −41.6% −61.9%
GLD (gold) 2005 8.0% 9.7% −30.2% −42.9%
TLT (long Treasuries, price) 2003 −0.8% −0.1% −32.0% −54.5%

The rule beat holding on return in QQQ and EEM, the two funds with the deepest falls, and trailed in the rest. It reduced the deepest fall in all six. The QQQ result depends on a start date at the top of the 2000 bubble. The TLT row is price only: the interest a bond fund pays, which is most of its return, is not in the data, so that row says little about bonds.

Compared with its neighbours

Rule on SPY Return per year Deepest fall Round trips
Buy and hold 8.9% −52.2% 1
Price above its 10-month average 7.8% −23.8% 25
12-month return above zero 8.5% −20.8% 14

The 12-month rule traded about half as often as the moving average rule and finished slightly ahead of it. The start months differ by three, so the rows are close but not identical tests. The two rules are members of one family, and a difference of this size between them is not evidence that one is better.

How Finabase does it

No Finabase tool runs this rule or tells you what to hold. These show its input:

  • Chart. The Rate of Change (ROC): percentage momentum and how to read it study is the rule's measurement: the percentage change over a chosen number of bars. On a weekly chart a period of 52 gives the 12-month return, and the rule is "above zero or not". The link above opens it on SPY.
  • Backtest. The Backtest tool trades the signals of the indicators you pick on one symbol. It differs from this page's test in two ways that change the numbers: each signal buys a fixed $10,000, and no trading cost is charged.
  • Strategy. The Strategy tool replays if/then rules over two years. This rule trades about once every two years, so the replay is too short to judge it.

Frequently asked questions

What is the difference between time-series and cross-sectional momentum?

Time-series momentum compares an asset with its own past: is it up over the last year? Cross-sectional, or relative, momentum compares assets with each other and holds the strongest, even if all of them are falling. The first can move to cash. The second is always invested.

Does time-series momentum beat buy and hold?

Not on return, in this SPY test: 8.5% a year against 8.9%. Its deepest fall was 20.8% against 52.2%. All of its advantage came from the bear markets of 2000 to 2002 and 2007 to 2009.

Why twelve months?

Twelve months is the lookback used in the original research and it trades least. In this test shorter lookbacks did worse. That is one history, and a test of four lookbacks cannot show that twelve is right in general.

Is this the same as the 200-day moving average?

No, but it is close. The two gave the same answer in 349 of 393 months on SPY. They differ at turning points, because one compares the price with an average and the other with a single price from a year ago.

Summary

Time-series momentum holds a fund when its close is above its close of twelve months earlier and holds cash otherwise. On 33 years of SPY it returned 8.5% a year against 8.9% for buying and holding, with a deepest fall of 20.8% against 52.2%. Half its trades lost; two long holdings supplied most of the gain. It was far ahead through the 2009 low and has trailed by 5.7 points a year since, losing most in sharp falls that reversed, such as December 2018 and March 2020. It is a close relative of the moving average rules and fails in the same markets.

Check your understanding

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

  1. A fund closes the month at $50. Twelve months ago it closed at $52. Every other fund in the market fell more. What does time-series momentum hold?

    1. The fund, because it is the strongest
    2. Cash, because the fund is below its own level of a year ago
    3. Half fund, half cash
    Show the answer

    B. Cash, because the fund is below its own level of a year ago Time-series momentum compares the fund only with its own past. Being stronger than other funds is relative momentum, a different rule.

  2. Half of the rule’s 14 trades on SPY lost money, yet it nearly matched buy and hold. How?

    1. The gains averaged 56.7% and the losses 2.5%
    2. The losing trades were not counted
    3. Costs were refunded on losing trades
    Show the answer

    A. The gains averaged 56.7% and the losses 2.5% The rule cuts a losing position within months and lets a winning one run for years. Two trades alone gained 169.3% and 81.5%. Win rate says little about a strategy shaped like this.

  3. The rule sold SPY in April 2022 at $412.00 and bought back in April 2023 at $415.93. What did it gain from being out during the 2022 decline?

    1. About a quarter, the size of the decline
    2. Nothing in price: it avoided the fall and missed the whole recovery
    3. The dividends
    Show the answer

    B. Nothing in price: it avoided the fall and missed the whole recovery Avoiding a fall only helps if the rule buys back lower. Here the market recovered before the 12-month return turned positive, so the rule had a calmer year and no extra return.

  4. Why can the rule’s answer change in a month when the price barely moved?

    1. Because the comparison price, from twelve months ago, changed
    2. Because trading costs changed
    3. It cannot
    Show the answer

    A. Because the comparison price, from twelve months ago, changed The rule compares today with one price a year back. When a crash month from last year becomes the comparison, the hurdle drops sharply, as it did in October 2009.

Sources and further reading

  1. Tobias J. Moskowitz, Yao Hua Ooi and Lasse Heje Pedersen, “Time Series Momentum,” Journal of Financial Economics 104(2), 2012 — the 12-month rule across 58 futures markets, 1985 to 2009
  2. Gary Antonacci, “Absolute Momentum: A Simple Rule-Based Strategy and Universal Trend-Following Overlay” (2013) — the single-asset form of the rule
  3. IRS — Topic no. 409, Capital gains and losses
  4. State Street — SPDR S&P 500 ETF Trust (SPY), fund page: inception January 22, 1993
  5. Finabase chart data — monthly bars for SPY (from February 1993), QQQ, IWM, EFA, EEM, GLD and TLT to October 2026; rules and results computed by Finabase