Markets › Stocks & ETFs · SPY ETF, daily, with dividends
200-Day Moving Average Strategy Backtest: SPY 1993-2026, With Dividends
Buy the S&P 500 above its 200-day moving average, sit in cash below it: the daily rule and Meb Faber's 10-month version on 33 years of SPY, with dividends, T-bill interest, costs and a fair comparison with chance.
Buried · FragileIt dodged the worst crashes. It never beat holding.
Variants tried: 2 · Passed every gate: 0 · Worst drawdown incl. open losses: -24.6% vs -55.2% buy-and-hold · Holdout opened: yes (already used, in part)
Short answer: as crash insurance, yes; as a way to beat the market, no. Holding SPY only while it closed above its 200-day moving average (T-bills otherwise) cut the worst drawdown, including open losses, from 55.2% to 24.6% from 1993 to 2026, dividends included, but returned 8.9% a year against 10.9% for buy-and-hold. Faber's monthly 10-month rule did slightly better (9.6% a year). Neither beat random holding periods convincingly.How much costs took here, compared across all 54 of our tests →
00The verdict in three lines
- We tested the 200-day moving average strategy on SPY from November 1993 to October 2026: be invested while SPY closes above its 200-day simple moving average, hold T-bills while it closes below, trade at the next open. Also Meb Faber's monthly 10-month rule. Dividends, T-bill interest and 0.04% a round trip included.
- It halved the worst crash and lost return. Worst drawdown, including open losses, -24.6% against -55.2% for buy-and-hold, but 8.9% a year against 10.9%: $1 grew to $16.37 instead of $29.77. Faber's monthly rule: 9.6% a year, worst drawdown including open losses -24.3%.
- Not proven as a signal. Random holding periods of the same lengths did almost as well per trade, and the daily rule's profits came from ten long trades. Buried (daily) and Inconclusive (monthly, only 25 trades).
01What is the 200-day moving average strategy?
The 200-day moving average is the average closing price of the last 200 trading days, about ten months. As a strategy it says: own the market while the price is above its 200-day moving average, step aside while it is below. It is one of the most quoted rules in investing. Paul Tudor Jones has called it his metric for everything he looks at, and Meb Faber's paper “A Quantitative Approach to Tactical Asset Allocation” (2007) made a monthly form of it famous: compare the month-end price with its 10-month average, hold Treasury bills when below.
- SMA-D200, the daily 200-day rule: at each close, compare SPY with the simple average of its last 200 closes. Above: buy at the next open (if not already in). Below: sell at the next open and hold T-bills. A close exactly on the average changes nothing.
- SMA-M10, Faber's 10-month rule: on the last trading day of each month, compare SPY with the average of its last 10 month-end closes; switch at the next day's open.
- The claim we tested: most of the market's return with much smaller crashes. We judge it both ways: against buy-and-hold, and with our six gates against chance.
02How we tested it
Data: SPY daily prices from its first day (29 January 1993) to 20 October 2026, 8,489 trading days, with all 136 dividends (Yahoo Finance history). Dividends count: while invested, the strategy earns them, and buy-and-hold is measured as total return. Many 200-day backtests use the price index alone, which makes time out of the market look cheaper than it is: over these years SPY's price rose about 17-fold, its total return about 32-fold. Cash: the 3-month T-bill rate (FRED) while out. Costs: 0.04% of the price a round trip; at 0.10% nothing changes materially. Signals use the closing price, trades the next open, so the rule never uses a price it could not know.
Periods: 1993 to 2005, 2006 to 2015, 2016 to 2026 (the last one is already used, in part: the S&P 500 from 2020 was in our golden-cross test on ES futures). Control: the same holding periods moved to random times, 500 times, to see whether the average picks better moments than chance in a market that mostly rises. Engine check: a self-test with hand-worked cases and an independent re-implementation that matched every trade (sma200_selftest_result.txt); in its first run one control case written in the test itself expected the wrong wrap-around, and was corrected before any result was printed.
03Does the 200-day moving average strategy work?
The whole strategy, cash interest included, against buy-and-hold SPY with dividends reinvested, 1993-11-11 to 2026-10-20:
| Strategy | $1 became | Return a year | Worst drawdown incl. open losses (end of day) | Volatility | Return per unit of volatility | Time invested | Weakest calendar year (closed at year end) |
|---|---|---|---|---|---|---|---|
| Buy and hold SPY | ×29.8 | 10.9% | -55.2% | 18.7% | 0.58 | 100% | 2008: -36.8% |
| Daily 200-day rule | ×16.4 | 8.9% | -24.6% | 11.9% | 0.74 | 75% | 2000: -19.0% |
| Faber 10-month rule | ×20.4 | 9.6% | -24.3% | 12.6% | 0.76 | 76% | 2022: -20.6% |
Growth of $1, log scale, from the first day both rules exist (11 November 1993).
Short answer: it works as crash insurance, not as a way to beat the market. The daily rule earned 2.0 percentage points a year less than buy-and-hold and ended with about half the money, but its worst fall, counting open losses at each close, was -24.6% instead of -55.2%, with about 64% of the volatility. Per unit of risk it did better (0.74 against 0.58); per dollar it did worse.
04How the 200-day rule did in every crash
Total return from the S&P 500's peak to its low in each bear market of the sample:
| Crash (peak to low) | Buy and hold | SMA-D200 | SMA-M10 |
|---|---|---|---|
| 2000-02 dot-com (2000-03-24 to 2002-10-09) | -47.5% | -14.9% | -5.2% |
| 2008-09 financial crisis (2007-10-09 to 2009-03-09) | -55.2% | -4.9% | -4.5% |
| 2020 Covid (2020-02-19 to 2020-03-23) | -33.7% | -18.9% | -11.9% |
| 2022 rate hikes (2022-01-03 to 2022-10-12) | -24.5% | -14.3% | -16.4% |
In the two slow bear markets, 2000-02 and 2008-09, both rules were out for most of the fall and lost a fraction of it. The Covid crash took five weeks, too fast for a 200-day average: the daily rule still lost 18.9%, the monthly one 11.9%. In 2022 the market fell in waves, and the rules were bought back on rallies and sold again; Faber's rule ended that calendar year at -20.6%, worse than buy-and-hold's -18.2%.
05Why it lags buy-and-hold
- Whipsaws. 69 of the daily rule's 112 trades lasted ten trading days or less; together they lost -86.4%. Each one is a sell near the average and a buy back a little higher.
- Late re-entries. After a crash the price must climb back above a falling average before the rule buys, so the first part of every recovery is missed: in 2009 the daily rule made +17.3% against +26.4%, in 2019 +14.9% against +31.2%, in 2023 +11.4% against +26.2%.
- Cash earns less than stocks. A quarter of the time out of the market, at T-bill rates, costs return in a market that rose most years.
The daily rule beat buy-and-hold in 7 of 34 calendar years (1997, 2001, 2002, 2008, 2013, 2018, 2022), the monthly one in 7 (1997, 2000, 2001, 2002, 2008, 2013, 2014). The large wins were the bear-market years 2001, 2002 and 2008; in years spent fully invested, such as 1997 and 2013, the difference is a few hundredths of a percent.
06Daily 200-day rule vs Faber's 10-month rule
| Variant | Trades | Winners (95% range) | Net per trade | Sum of trade returns | Profit factor | Median days held | Sum without the 10 best | Verdict |
|---|---|---|---|---|---|---|---|---|
| SMA-D200 | 112 | 35.7% (27% to 45%) | +2.87% | +320.9% | 3.71 | 6 | -18.1% | Buried |
| SMA-M10 | 25 | 64.0% (45% to 80%) | +14.57% | +364.2% | 8.15 | 189 | -19.9% | Inconclusive |
The monthly check trades far less (25 trades in 33 years against 112), wins more often (64% against 36%) and ended with more money (×20.4 against ×16.4), because it ignores most of the daily crossings. It also reacts later: in 2022 that cost it more. Trade returns include dividends and costs; their sum is a simple count, not a compounded return.
07Year by year
| Year | Buy and hold | SMA-D200 | SMA-M10 |
|---|---|---|---|
| 1993 (from 11 Nov) | +1.1% | +0.9% | +1.1% |
| 1994 | +0.4% | -4.5% | -7.6% |
| 1995 | +38.0% | +37.0% | +34.0% |
| 1996 | +22.5% | +17.1% | +22.5% |
| 1997 | +33.5% | +33.5% | +33.5% |
| 1998 | +28.7% | +16.9% | +11.9% |
| 1999 | +20.4% | +9.8% | +13.3% |
| 2000 | -9.7% | -19.0% | +0.4% |
| 2001 | -11.8% | +3.5% | +3.5% |
| 2002 | -21.6% | -2.6% | -4.1% |
| 2003 | +28.2% | +25.0% | +23.0% |
| 2004 | +10.7% | +3.5% | +7.4% |
| 2005 | +4.8% | +1.6% | +2.0% |
| 2006 | +15.8% | +9.8% | +15.8% |
| 2007 | +5.1% | +2.2% | +5.1% |
| 2008 | -36.8% | +1.0% | +1.6% |
| 2009 | +26.4% | +17.3% | +22.0% |
| 2010 | +15.1% | -0.2% | +2.2% |
| 2011 | +1.9% | -6.3% | -1.9% |
| 2012 | +16.0% | +13.9% | +10.0% |
| 2013 | +32.3% | +32.3% | +32.3% |
| 2014 | +13.5% | +11.1% | +13.5% |
| 2015 | +1.3% | -9.5% | -6.7% |
| 2016 | +12.0% | +11.1% | +9.4% |
| 2017 | +21.7% | +21.7% | +21.7% |
| 2018 | -4.6% | -3.7% | -7.4% |
| 2019 | +31.2% | +14.9% | +6.8% |
| 2020 | +18.4% | +8.5% | +15.6% |
| 2021 | +28.7% | +28.7% | +28.7% |
| 2022 | -18.2% | -15.1% | -20.6% |
| 2023 | +26.2% | +11.4% | +10.9% |
| 2024 | +24.9% | +24.9% | +24.9% |
| 2025 | +17.7% | +11.6% | +12.4% |
| 2026 (to 20 Oct) | +13.7% | +10.7% | +3.3% |
Total return with dividends and T-bill interest. Green: better than buy-and-hold that year.
08Is it the average or chance?
Our protocol asks whether the signal picks better times to be invested than the same holdings placed at random. The control moves every real holding period, all together, by a random number of trading days (500 times), keeping its length. In a market that rises most of the time, random holdings make money too.
| Variant | Trades | Real net per trade | Control median | Control 5th to 95th | Percentile (95 needed) |
|---|---|---|---|---|---|
| SMA-D200 | 112 | +2.87% | +2.47% | +1.89% to +3.16% | 81.6 |
| SMA-M10 | 25 | +14.57% | +12.01% | +8.17% to +15.59% | 88.4 |
By period
| Variant | 1993 to 2005 (trades) | 2006 to 2015 (trades) | Holdout* 2016 to 2026 (trades) | t, first period | PF, second period |
|---|---|---|---|---|---|
| SMA-D200 | +123.1% (45) | +71.2% (36) | +126.6% (31) | 1.42 | 2.81 |
| SMA-M10 | +179.8% (9) | +75.5% (5) | +108.9% (11) | 1.56 | 10.49 |
The gates
| Variant | Trades | 1. PF ≥ 1.2 in validation | 2. ≥ 50 trades in each period | 3. t ≥ 2.0 on development | 4. Above the 95th percentile of the shifted-holding control | 5. Positive without the 10 best trades | 6. Holdout net positive (partly seen) |
|---|---|---|---|---|---|---|---|
| SMA-D200 | 112 | pass | fail | fail | fail | fail | pass |
| SMA-M10 | 25 | pass | fail | fail | fail | fail | pass |
The daily rule was positive in all three periods, but with 45, 36, 31 trades (50 needed in each), a first-period t of 1.42 (2.0 needed), a control percentile of 81.6 (95 needed) and a loss without its ten best trades. Cause: Fragile (positive, but not without its ten best trades). The monthly rule has fewer than 30 trades, so by our rules it is Inconclusive, whatever its numbers. * Partly seen before this test.
TWhat this shows, and what to take into your own tests
What the results show
- Holding SPY only above its 200-day average, and T-bills below it, cut the worst drawdown, counting open losses at each close, from -55.2% to -24.6% from 1993 to 2026, and stepped around most of the 2000-02 and 2008-09 bear markets.
- It paid for that with return: 8.9% a year against 10.9% for buy-and-hold with dividends, ×16.4 against ×29.8. It beat buy-and-hold in only 7 of 34 calendar years.
- As a trading signal it is not proven: the same holding periods placed at random times did almost as well per trade (percentile 82), and without its ten best trades the daily rule lost money.
What remains unknown
- How it does in the next bear market. Two of the four crashes here were slow and the rule shone; the fast ones (2020) and the choppy ones (2022) went much less well.
- Taxes. Every sell in a taxable account realises gains; buy-and-hold defers them. That would widen the gap.
- Other assets and filters: QQQ, TQQQ, a band around the average, or a few days of confirmation. We tested the plain rules with their usual settings only.
Why it was buried
Our gates judge a rule by its trades, and these trades depend on a handful of big winners: 69 of the daily rule's 112 trades lasted ten days or less and lost -86.4% together, while ten long holdings made all the money. With 45, 36, 31 trades in our three periods, the evidence is too thin, and random holdings of the same lengths came close. The monthly rule traded only 25 times in 33 years, too few for any verdict.
What you can apply to your own tests
- Decide what you want from a trend filter before you use one. It is insurance against long bear markets, and insurance costs money in the years without one.
- Count the whipsaws. A daily rule on the 200-day average sold and bought back within two weeks in most of its trades.
- Compare with buy-and-hold including dividends. Price-only backtests flatter any rule that spends time out of the market.
A note on the numbers: drawdown figures are measured on closed trades unless a table says it includes open losses. Open losses inside a trade are not counted in a closed-trade drawdown, so real account risk is higher, especially for systems with no stop. Backtests are not live trading.
FAQCommon questions
Does the 200-day moving average strategy work?
It reduces crashes but costs return. On SPY from 1993 to 2026 with dividends and T-bill interest, the daily rule returned 8.9% a year with a worst drawdown (including open losses) of 24.6%; buy-and-hold returned 10.9% with a worst drawdown of 55.2%, also including open losses. It beat buy-and-hold in 7 of 34 calendar years, mostly bear-market years.
What is the 200-day moving average strategy?
Own the market (here SPY) while its closing price is above the average of its last 200 closes, and hold cash or T-bills while it is below. We bought or sold at the next open after each crossing.
Is the 10-month moving average better than the 200-day?
In our test Faber's monthly 10-month rule traded 25 times in 33 years against 112 for the daily 200-day rule, ended with more money (20.4 times the start against 16.4) and had a similar worst drawdown including open losses (24.3%), but it reacted later in 2022.
How did the 200-day moving average do in market crashes?
It avoided most of the 2000-02 and 2008-09 bear markets (4.9% loss in 2008-09 against 55.2% for buy-and-hold), but lost 18.9% in the five-week Covid crash and 14.3% in 2022.
Why does the 200-day moving average lag buy-and-hold?
Whipsaws and late re-entries: 69 of the daily rule's 112 trades lasted ten days or less and lost money together, and after each crash the rule buys back only once the price is above a falling average, missing the start of the recovery.
How was the 200-day moving average tested?
On SPY daily prices from 1993 to 2026 with dividends, with rules written and dated before the data was run, costs included on every trade, results split into development, validation and holdout periods, a random-entry control as the benchmark, and every variant reported. Read the protocol.
09Check our arithmetic
sma200_spec.md
sma200_lib.js · sma200_ref.js · sma200_selftest.js · sma200_run.js · spy_parse.js
control_audit_lib.js · gh_lib.js · hourly_lib.js · ma_lib.js · rsi2_protocol.js
dtb3_fred_raw.csv
sma200_selftest_result.txt · sma200_results.txt · sma200_summary.json
Code provenance: our own code, written from the rules in the specification.
10Data notes
Source. Yahoo Finance daily history of SPY (prices adjusted for splits, which SPY never had; dividends listed separately), read from the saved history page; our close-plus-dividends index agrees with Yahoo's adjusted close to 0.05% over 33 years. Gaps. Only real market closures (September 2001, 2 January 2007, October 2012). Not included: taxes, and SPY's 0.09% yearly fee is already inside its price.
11Corrections
22 Oct 2026, navigation: added a link to the new guide comparing trading costs across all our tests. No result changed. 22 Oct 2026, navigation: added a link to the new guide comparing trend following and mean reversion across all our tests. No result changed.
+More tests and guides
- Golden cross backtest
- Macd crossover strategy backtest
- Supertrend indicator backtest
- Donchian channel breakout strategy backtest
- Trend following vs mean reversion: this test next to the others of its style
- Free prop firm challenge calculator
- What all the backtests found so far
- The backtesting protocol
- How to backtest a trading strategy without fooling yourself
- Backtesting glossary