Moving-average trading: the 200-day line, golden cross and 10-month rule

This article covers rules that switch between stocks and cash using one or two moving averages. For each rule it gives the exact definition, the variants people use, results on QQQ and SPY since 2000, and the stretches where the rule won and lost. Figures are as of September 2026. The golden cross results are recalculated every month on the golden cross backtest page in Questions and answers. The other rules were calculated separately by the operator with the same engine.

Test conditions

$10,000 went in once, with nothing added later. Signals are read on split-adjusted closes, and the whole account buys or sells at the next trading day's open. Fills and valuation use dividend-reinvested prices, and each buy or sell costs 0.15%. Cash earns no interest. The period runs from January 3, 2000 to September 30, 2026, and the benchmark is buy and hold with the same money on the same day.

Rules and variants

• Price against one line: when the close is above its n-day simple moving average, the rule buys at the next open and holds. When the close is below, it sells at the next open and holds cash. The 200-day line is the most common choice in the US. In Korea the 120-day line is often called the business-cycle line and the 60-day line the supply-and-demand line.
• 10-month rule: the rule is invested when the last close of the month is above the average of the last 10 month-end closes, and in cash when it is below. It is known from Mebane Faber's paper in the Spring 2007 issue of the Journal of Wealth Management (SSRN). The paper fills at the month-end close and pays 90-day Treasury bill interest on cash. Our calculation fills at the next trading day's open and pays no interest.
• Golden cross and death cross: a golden cross is the day the 50-day line crosses above the 200-day line, and a death cross is the day it crosses below. Korean charts often watch the cross of the 20-day and 60-day lines as well. Our calculation uses the day of the cross as the signal. If the 50-day line is already above on the start date, the rule judges that day's close and buys at the next trading day's open, as the Questions and answers page does.
• Confirmation filters: trading every time price crosses the line produces many whipsaws. So some rules require the close to stay beyond the line for n days in a row, and some put a percentage band (an envelope) around the line. Jeremy Siegel tested the Dow by buying on a close at least 1% above the 200-day line and selling on a close at least 1% below it (as cited in Faber's paper).
• Weekly reference lines: the 120-week and 200-week lines are watched less as trading signals than as a gauge of how far a long decline has gone. They have their own section below.

Results (as of September 2026)

Every rule had a smaller max drawdown than buy and hold. Annual returns depended on the ETF and the rule. On QQQ the 50/200 cross returned 11.2% a year against 8.6% for holding, and the 10-month rule and the 200-day line with a 1% band were close to holding. On SPY every rule we ran trailed buy and hold (8.2% a year). Shorter lines did worse. The 120-day rule returned 4.4% a year on QQQ and 3.0% on SPY, and the 20/60 cross returned 4.3% on QQQ and 4.8% on SPY.

The 50/200 numbers on QQQ depend on how the start date is handled. On January 3, 2000 the 50-day line was already above the 200-day line, and this article's calculation bought right away and sold at the September 2000 death cross at about the same price. The rule's max drawdown of −36.5% came during that first holding, between March and May 2000. Waiting for the first golden cross, as the custom rule in the backtest tool does, kept the rule in cash until January 2003, with 11.1% a year and a max drawdown of −28.6%. The annual return was about the same, and the max drawdown was 7.9 points smaller.

Where the rules won and where they lost

By decade, 2000 to 2009 was the only stretch where the rules beat holding. The QQQ 50/200 cross returned 5.0% a year in that decade against −6.9% for holding, and 12.1% in 2010 to 2019 against 17.9%. The QQQ 200-day rule returned 0.8% and 8.0% a year in the same two decades. SPY's 120-day rule trailed even in 2000 to 2009, at −2.6% a year against −1.1% for holding.

• A long bear market, 2000 to 2002: QQQ fell 83.0% from its March 27, 2000 closing high to October 9, 2002. The 50/200 cross sold at the September 2000 death cross and stayed in cash until January 2003. The 200-day rule climbed back above the line on every rally and made 15 round trips through early March 2003, 13 of them losses. Its max drawdown, −55.6%, ran from March 2000 to March 2003.
• The financial crisis, 2008 to 2009: QQQ fell 53.4% with dividends from October 31, 2007 to November 20, 2008. The 200-day rule sold on January 7, 2008, went through 8 short round trips from May to September that year, and bought back on May 1, 2009 at a price 28.8% below its sale. The 50/200 cross sold on February 4, 2008 and bought back on June 2, 2009, 20.2% lower.
• A V-shaped rebound, 2020: QQQ fell 28.6% from its February 19 closing high to March 16 and then recovered quickly. The 200-day rule sold on March 12 and bought back on April 9 at a price 11.4% above its sale. The 50/200 cross took the whole decline, sold on April 30 while the rebound was under way, and bought back on May 22, 3.9% higher. Its account fell 28.6% from its peak in this stretch.
• The 2022 bear market: QQQ fell 35.1% with dividends from its December 2021 high to its November 2022 low. The 200-day rule sold on January 21, 2022 and bought back on March 14, 2023 at a price 17.5% below its sale, but six short round trips in between left the account 10.0% lower over that span. QQQ fell 16.7% with dividends over the same span. The 50/200 cross sold on March 2, 2022 and bought back on March 14, 2023, 13.6% lower.
• A steady bull market, the 2010s: when declines are short and shallow, the rule keeps selling and buying back higher. From 2010 to 2019 the QQQ 200-day rule returned 8.0% a year, less than half of holding (17.9%).

Lag: the lines move late

An n-day simple moving average trails a steadily moving price by (n−1)/2 days: about 100 days for the 200-day line and about 25 for the 50-day line. A cross comes only after both lines have turned, so it comes later still. Since 2000, QQQ's 17 death crosses came a median 10.9% below the prior high and a median 58 trading days after it. Its 17 golden crosses came a median 12.2% above the prior low and 55 trading days after it.

So a cross rule gains only in declines much deeper than the drop it takes to trigger the signal (about 10%). It gained in 2000 to 2002 and in 2008, when the market fell more than 50%. In 2020, when the market hit bottom within a month, it took the whole decline and then bought back higher.

Whipsaws: the cost of trading near the line

When price moves back and forth around the line, trades pile up. The QQQ 200-day rule made 87 round trips, and 60 of them ended within a month (21 trading days). Only 30% of its trades ended in a gain, but the average winner (+16.8%) was much larger than the average loser (−2.1%), and 7.9% a year remained. For the 120-day rule, 104 of 149 round trips ended within a month. The 50/200 cross had only 2 of 18 end within a month, 67% of its trades won, and the average winner gained +31.7%. The 20/60 cross had 12 of 62 end within a month.

Confirmation filters: bands and closes in a row

A 1% band around the 200-day line cut fills from 174 to 88 on QQQ and from 190 to 82 on SPY, and raised annual returns from 7.9% to 8.8% on QQQ and from 5.5% to 6.5% on SPY. Requiring three closes in a row also beat the plain rule, at 8.4% a year on QQQ and 7.1% on SPY. But QQQ returned 8.8%, 7.8% and 8.8% a year with bands of 1%, 2% and 3%, and going from three days to five lowered the return from 8.4% to 7.8%. Wider was not steadily better, so the single best setting may simply fit this period.

The 120-week and 200-week lines

A weekly close (the close on the last trading day of the week) below the 120-week or 200-week line is rare. The figures below assume a purchase at the close of the first week below the line and are total returns with dividends. A new drop below the line within 13 weeks counts as the same episode.

Since September 1997, when its 200-week line can first be calculated, SPY has closed below it six times.

• March 16, 2001: +1.3% after one year, −0.4% after three. The weekly close fell another 29.7% after that.
• March 7, 2008: −46.2% after one year, +9.2% after three.
• June 20, 2008: −30.3% after one year, +5.0% after three. The weekly close fell another 47.6% after that.
• August 19, 2011: +28.9% after one year, +88.3% after three.
• March 20, 2020: +74.3% after one year, +79.4% after three.
• September 30, 2022: +19.9% after one year, +95.2% after three.

The 200-week line did not mark the bottom. In 2001 and 2008 the market fell another 29.7% and 47.6% after the first close below it, while in 2011, 2020 and 2022 that week was near the low. QQQ data start in March 1999, so its 200-week line can only be calculated from January 2003, when QQQ was already 54% below it. Since then it has closed below the line three times, in September 2008, July 2010 and September 2022, with one-year total returns of +3.8%, +37.4% and +33.7%. The 120-week line broke more often: 9 times for SPY and 7 times for QQQ (leaving out the stretch when QQQ was already 39% below the line when it was first calculated in June 2001). The one-year return was negative twice for SPY and twice for QQQ. A quote in which Charlie Munger supposedly talks about buying quality stocks near the 200-week line circulates widely, but its original source has not been found.

Adding interest on cash

The results above pay nothing on cash. From January 2000 to September 2026 the US 3-month Treasury bill rate (^IRX) averaged 1.9% a year. When the operator separately recalculated the same trades with that rate paid on cash, the QQQ 50/200 cross rose from 11.2% to 11.8% a year, and the SPY 50/200 cross rose from 7.9% to 8.4%, above SPY buy and hold (8.2%). The 200-day rule rose from 7.9% to 8.5% on QQQ and from 5.5% to 6.1% on SPY. The gap between rule and holding is small enough that interest on cash alone can reverse it.

Related concepts and search terms

Moving-average rules go by many names, but they belong to the same family. Here are the names to use when you look further.

• Trend following: the whole family of methods that hold rising assets and avoid falling ones. Moving-average rules are its simplest form, and Turtle trading is an example of the same idea built into a full system for futures markets.
• Time-series momentum: going long an asset when its own past-year return is positive and short when it is negative. Moskowitz, Ooi and Pedersen named it in a 2012 study of futures markets. In a long-only form it belongs to the same family as rules that check whether price is above its moving average.
• Absolute momentum: Gary Antonacci's name for one half of dual momentum. It holds stocks only when their 12-month return is above the return of short-term Treasury bills, and otherwise moves to bonds.
• Bullish and bearish alignment: lines stacked from shortest at the top to longest at the bottom, or the reverse. The 50/200 cross uses the days when this alignment of the two lines starts and ends as signals.
• Moving average envelope: a band drawn a fixed percentage above and below a moving average. The 1% band rule in this article is one.
• 200-day rule with monthly investing: results for the same 200-day rule while putting money in every month are on the 200-day rule backtest page.

Search terms: moving average crossover, golden cross backtest, death cross, 200-day moving average strategy, 10-month SMA timing, trend following, time series momentum, whipsaw, moving average envelope.

Drawbacks and risks

• Every sale has costs and taxes. The calculation includes only the 0.15% fee and leaves out capital gains tax. The QQQ 200-day rule made 87 round trips after 2000.
• Results are sensitive to the start date and the period. The QQQ 50/200 cross earned its excess return in 2000 to 2009, and its max drawdown was −36.5% or −28.6% depending on how the start date was handled.
• The more settings you choose, the more the result fits the past. Band widths and confirmation days did not move results in one direction.
• Signals were read at the close and filled at the next open. On days with a large opening gap, the fill price differs from the price that produced the signal.
• These results come from index ETFs. There is no basis for expecting the same numbers from individual stocks that never recover.