Build and test a strategy Lesson 7 of 14
Strategies · Trend and momentum
Moving average trend following: the 10-month rule, tested
The 10-month moving average rule on SPY from 1993 to 2026: the exact rules, a backtest with costs, what it avoided in 2008 and what it gave up afterwards.
On this page
Trend following with a moving average is one of the simplest strategies there is: own the market while its price is above its own average, and step aside when it falls below. The best known version checks once a month and uses a 10-month simple moving average. Mebane Faber described it in a 2007 paper, and it has been argued over ever since.
This lesson states the rule exactly, tests it on 33 years of SPY, the oldest S&P 500 exchange-traded fund, and shows both halves of the result. The rule cut the two worst falls of the period to a fraction. It also earned less than simply holding, and since 2009 it has earned far less. Both facts follow from the same mechanism, and seeing why is the point of the lesson.
The idea
Markets fall in two ways. Some falls are short: a month or two down, then straight back. Others grind on for a year or more, like 2000 to 2002 and 2007 to 2009. A long decline has to pass through "below its recent average" on the way down, and it stays there for most of the trip.
A moving average rule uses that. It cannot know which kind of fall has started, so it sells in both. In a long decline that is a large saving. In a short one the price recovers, the rule buys back, and the round trip usually costs a few percent. The strategy is a trade: many small losses in exchange for missing the middle of the big declines.
Notice what the rule does not claim. It does not forecast. It gets out after the price has already fallen and gets back in after it has already recovered. It will never sell the top or buy the bottom.
The rules
Checked once a month, on the last trading day:
- Take the fund's closing price on the last trading day of each of the last ten months, including this one, and average them.
- If today's close is above that average, hold the fund for the next month.
- If it is at or below the average, hold cash for the next month.
- Do nothing between month-ends, whatever happens.
That is all of it. There are no stops, no targets and no judgment calls, which is what makes it testable. A rule a person can bend cannot be backtested.
The shaded bands in the figure are the months in cash. Two are wide: most of October 2000 to April 2003, and January 2008 to June 2009. The rest are narrow, and most of the narrow ones were mistakes.
A worked example
At the end of December 2007 SPY closed at $146.21. The average of its last ten month-end closes was $148.95. The close was below the average, so the rule moved to cash.
It stayed in cash for 18 months. At the end of June 2009 SPY closed at $91.95, above its 10-month average of $90.13, and the rule bought. It had sold at $146.21 and bought back at $91.95, 37% lower. The lowest month-end close in between, February 2009, was lower still: the rule missed the bottom by four months and still came out far ahead.
What the backtest shows
The test, stated in full so it can be checked:
- Data. SPY month-end closes from February 1993 to October 2026. The first ten months build the first average, so the test runs from November 1993: 396 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, to stand for the spread and any commission.
- Left out. Dividends, interest on cash and taxes. The first two partly offset: buy and hold collects every dividend, while the rule collects them about three-quarters of the time and would have earned interest on cash the rest.
| Measure | 10-month rule | Buy and hold |
|---|---|---|
| Total return | 1,099.5% | 1,530.8% |
| Return per year (CAGR) | 7.8% | 8.9% |
| Deepest fall from a peak | −23.8% | −52.2% |
| Longest time below a peak | 37 months | 80 months |
| Time invested | 76% | 100% |
| Round-trip trades | 25 | 1 |
| Trades with a gain | 64% | n/a |
One dollar became $12.00 under the rule and $16.31 held throughout. The rule gave up one percentage point a year and, in return, never lost a quarter of its value. Buy and hold lost half, twice.
The costs matter less than might be expected because the rule trades so little. With no cost at all it returns 8.0% a year, so 49 trades at 0.10% took about 0.2 points a year.
The average winning trade gained 22.6% and the average losing trade lost 5.3%. That shape, a few large gains and many small losses, is typical of trend following. It also means the win rate says little on its own.
The two halves of the period
Splitting the test at the February 2009 low changes the story completely.
| Period | 10-month rule | Buy and hold |
|---|---|---|
| November 1993 – February 2009 | 8.0% a year | 3.1% a year |
| February 2009 – October 2026 | 7.7% a year | 14.1% a year |
Through 2009 the rule was far ahead, because the period contained two long bear markets. From August 2000 to September 2002 buy and hold fell 46.3% and the rule fell 11.5%. From October 2007 to February 2009 buy and hold fell 52.2% and the rule 5.6%.
Since 2009 there has been no decline that lasted, and the rule has earned a little over half of what holding earned. Anyone who read the 2007 paper and started the rule in 2009 has paid for protection for 17 years without the event it protects against.
Where it fails: the losing periods
The rule loses when the market falls far enough to cross its average and then turns straight back up. The rule sells near the low and buys back higher. This is called a whipsaw. The table lists the clearest cases since 2018.
| Sold | Bought back | Result |
|---|---|---|
| December 2018, $249.92 | February 2019, $278.68 | Bought back 11.5% higher |
| February 2020, $296.26 | May 2020, $304.32 | Missed the March fall; back 2.7% higher |
| April 2022, $412.00 | November 2022, $407.68 | About even, then sold again a month later |
| October 2023, $418.20 | November 2023, $456.40 | Bought back 9.1% higher |
| March 2025, $559.39 | May 2025, $589.39 | Bought back 5.4% higher |
| March 2026, $650.34 | April 2026, $718.66 | Bought back 10.5% higher |
From December 2019 to October 2026 the rule gained 63.1% and buy and hold gained 134.8%. The rule's own deepest fall, 23.8% from December 2021 to February 2023, came from this kind of period: it bought in March 2022 and sold in April about 9% lower, then bought in November 2022 and sold in December about 6% lower.
Other ways it fails or misleads:
- A fall inside one month. The rule looks once a month. A crash that begins and ends between two month-ends is taken in full.
- Taxes. In a US taxable account every sale at a gain is taxed in that year, and gains on positions held a year or less are taxed as ordinary income (IRS Topic 409). Buy and hold defers all of it. The test ignores this, and it counts against the rule.
- Sticking with it. The rule trailed for most of 17 years. A strategy only produces its backtest for someone who follows it through the stretch when it looks broken.
- One market. This is the US stock market over a period that ended very well. Faber's paper applied the rule to five asset classes for that reason.
Is ten months special?
No, and that is reassuring rather than worrying. Changing the length changes the detail but not the shape of the result.
| Average length | Return per year | Deepest fall | Round trips |
|---|---|---|---|
| 6 months | 6.9% | −27.7% | 38 |
| 8 months | 7.7% | −23.8% | 31 |
| 10 months | 7.8% | −23.8% | 25 |
| 12 months | 8.3% | −18.7% | 21 |
Every length trailed buy and hold's 8.9% and every length roughly halved the deepest fall. A result that held for only one setting would be a sign of overfitting. The 12-month line looks best here, which is exactly the kind of finding not to trust: it is the best of four tries on one history.
Compared with its neighbours
| Approach | Trades | Deepest fall in this test | What it asks of the investor |
|---|---|---|---|
| Buy and hold | None | −52.2% | Sitting through a fall of half |
| 10-month rule | About one a year | −23.8% | Following signals that are often wrong |
| 50-day and 200-day crossover | Similar, on daily bars | Not tested here | The same trade-off, checked daily |
| Holding stocks and bonds together | Rebalancing only | Not tested here | Accepting a lower return in every year |
The crossover of the 50-day and 200-day averages, covered in Moving averages and trend: filters, the 50/200 and golden crosses, is the same idea on daily data. Ten months is about 210 trading days, so the two rules tend to disagree by weeks, not by direction.
How Finabase does it
Finabase does not run this monthly rule for you, and no Finabase tool tells you when to buy or sell. Three tools let you look at the pieces.
- Chart. The chart draws the simple moving average on daily bars. A 200-day SMA on SPY is the closest daily equivalent of the 10-month average, and the link above opens it.
- Backtest. The Backtest tool trades every signal of the indicators you pick on one symbol. It differs from the test on this page in two ways that change the numbers: each signal buys a fixed $10,000, and it charges no trading cost.
- Strategy. The Strategy tool takes if/then rules, such as "if SPY is above its average, hold SPY", and replays them over two years. Two years is too short to judge a rule that trades once a year; use it to see how a rule behaves, not whether it works.
Frequently asked questions
Does the 10-month moving average rule beat the market?
Not on return, in this test. On SPY from 1993 to 2026 it earned 7.8% a year against 8.9% for buying and holding. What it changed was the size of the worst fall: 23.8% against 52.2%.
Why use month-end prices and not daily ones?
Checking once a month means fewer trades and fewer false signals. A daily check of the 200-day average on the same fund crosses back and forth many more times in a choppy market, and each crossing costs money. The price of checking monthly is that a fast crash inside a month is not avoided.
What does the rule hold when it is out of the market?
Cash, in this test, earning nothing. Faber's paper used Treasury bills, which would have added to the rule's return in the years when short-term rates were well above zero. That is one reason this test is not a copy of the paper's results.
Is the rule the same as the 200-day moving average?
Nearly. Ten months is about 210 trading days. The 200-day rule is usually checked every day, so it reacts sooner and trades more often.
Summary
The 10-month rule holds a fund while its month-end close is above its 10-month average and holds cash otherwise. Over 33 years of SPY it earned a point a year less than buying and holding and roughly halved the worst fall. All of its advantage came from two long bear markets before 2009; since then it has mostly paid for protection it did not need, through a string of sells that were followed by higher prices. Whether that trade is worth making depends on what a fall of half would do to the person holding the portfolio, which no backtest can answer.
Check your understanding
4 questions on this lesson. Each answer comes with the reason, and nothing is scored.
At a month-end a fund closes at $98. Its last ten month-end closes, including this one, average $100. What does the 10-month rule hold next month?
- The fund, because the price is close to the average
- Cash, because the close is below the average
- Half of each
Show the answer
B. Cash, because the close is below the average The rule has no middle setting and no judgment: a close at or below the average means cash for the next month, and the question is asked again at the next month-end.
In the SPY test the rule returned 7.8% a year and buy and hold 8.9%. What did the rule give in exchange for the lower return?
- A higher share of winning trades
- A much smaller worst fall: 23.8% against 52.2%
- Lower trading costs
Show the answer
B. A much smaller worst fall: 23.8% against 52.2% The rule was in cash for most of 2000 to 2003 and 2008 to 2009, so its deepest fall was less than half of buy and hold’s. It traded more, not less, and its win rate says little because its gains were much larger than its losses.
Which market does the most damage to a moving average rule?
- A decline that lasts two years
- A steady rise with no pullbacks
- A sharp fall that recovers within a few months
Show the answer
C. A sharp fall that recovers within a few months A quick fall and recovery makes the rule sell low and buy back higher, a whipsaw. In December 2018 it sold SPY at $249.92 and bought back two months later at $278.68. A long decline is what the rule is built for.
Averages of 6, 8, 10 and 12 months were all tested, and 12 months had the best return. What is the sound conclusion?
- Twelve months is the right setting from now on
- The result is similar across lengths, so it does not depend on one lucky setting; the best of four tries is not evidence
- The rule does not work, because the settings disagree
Show the answer
B. The result is similar across lengths, so it does not depend on one lucky setting; the best of four tries is not evidence All four lengths trailed buy and hold on return and all roughly halved the worst fall. That stability is the useful finding. Picking the winner of several tries on one history is overfitting.
Exercise
One task to do in Finabase. Nothing is graded or saved.
Open the two-year SPY chart with the 200-day SMA, the closest daily stand-in for the 10-month average. Find each stretch where the price closed below the line. For each one, note roughly how long it stayed below, and whether the price was higher or lower when it came back above the line than when it first went under. Which stretches would have helped a rule that sells below the line and buys back above it, and which cost it?
Open the SPY chart with the 200-day SMAWhat you should find
A stretch helps only if the price comes back above the line lower than where it went under. A short dip that returns higher is a whipsaw: the rule sells, then pays more to get back in. The lesson's table of cases since 2018 is made of these. Long stays below the line with a much lower re-entry, as in 2000 to 2002 and 2008, are what the rule exists for, and a two-year window may hold none.
Sources and further reading
- Mebane T. Faber, “A Quantitative Approach to Tactical Asset Allocation,” Journal of Wealth Management (Spring 2007) — the 10-month moving-average timing rule
- Jeremy J. Siegel, Stocks for the Long Run, 5th ed. (McGraw-Hill, 2014) — tests of the 200-day moving-average rule
- IRS — Topic no. 409, Capital gains and losses
- Finabase chart data — SPY monthly bars, February 1993 to October 2026; rule and results computed by Finabase
