Over 32.9 years of SPY (the S&P 500 ETF), buy and hold returned 10.84% a year. None of the ten moving average crossover pairs we tested matched it. They returned between 4.49% and 9.93% a year.
Which pair works best? Our earlier test picked each year's pair using only earlier data, across a much wider set of tickers. The best pair did not persist. Choosing the past best pair did no better than choosing at random. This guide asks what the whole family of crossover rules does to return and risk.
What is a moving average crossover? A crossover rule holds a fund while a fast moving average is above a slow one. Otherwise it holds cash.
Across 105 tests on seven funds, between 5 and 12 rules beat buy and hold on return, depending on where the window starts. Between 82 and 100 had a shallower worst drawdown, the largest fall from a peak to a later low.
The drawdown count is 100 in the two windows that open before 2008, which contain the 2008 crash in full. It is lowest in the window that opens in October 2015.
On each fund's full history, ten of the 105 rules left the worst drawdown deeper than holding, two by near ties. The 10-day length did so on five of seven funds.
How a moving average crossover works
A moving average is the mean of a fund's recent closing prices, as What Is a Moving Average? explains. A crossover rule uses a fast average over fewer days and a slow one over more. It holds the fund while the fast one is above the slow one. Otherwise it holds cash.
A pair is written fast/slow, so 50/200 pairs a 50-day average with a 200-day average. We tested the ten pairs drawn from the lengths 10, 20, 50, 100 and 200 days. We also tested each length alone, holding the fund while its price closes above that average.
The seven funds were SPY, QQQ (the Nasdaq-100 ETF), DIA (the Dow Jones Industrial Average ETF), IWM (the Russell 2000 ETF), GLD (the gold ETF), SLV (the silver ETF) and TLT (the long-term US Treasury bond ETF).
Each signal is read at the close. The position then earns the next day's move. Cash earns nothing. Each change of position costs 5 basis points, or five hundredths of a percentage point.
Do moving average crossovers beat buy and hold?
The headline comes from 105 tests: 15 rules on each of seven funds. We repeated them with the window opening on seven dates. The first, 28 April 2006, is SLV's first trading day, an accident of instrument history.
Source: YX Insights
Chart 1 plots both counts. The number of rules beating buy and hold on return stayed between 5 and 12. The number with a shallower worst drawdown stayed between 82 and 100. Ties with buy and hold make the counts between the endpoints sensitive to rounding, so only the range is shown.
In these tests, moving averages rarely added return. They usually cut the worst fall. These are not 105 independent tests, because four of the funds are views of one market. The counts are not a hit rate.
Why the drawdown result depends on the 2008 crash
The two windows that open before 2008 each return 100 of 105 on drawdown. Both sit through the 2008 crash in full.
Every later window returns less. Each begins only once its 200-day average exists, so the window nominally opening in January 2008 starts in October 2008, after most of that year's fall. The last starts in October 2015. From 2009 the drawdown that binds is the 2020 fall, not the 2008 crash.
On SPY, buy and hold's worst drawdown, 55.2%, came in the slow 2007 to 2009 fall. The 50/200 pair's loss in that fall stayed smaller, but its own worst, 33.7%, came in 2020, when it was invested through all 24 trading days of a fast one.
When a crossover rule makes the worst case worse
On full histories, ten rules left the worst drawdown deeper than buy and hold. Seven were single lengths. Three were silver pairs.
Two of the ten are near ties. The 10-day single length did it on five of the seven funds, QQQ's by a near tie. On SPY its worst drawdown was 64.7% against 55.2%, over 1,464 changes of exposure.
Trade count alone does not explain it. The 20-day rule changed exposure 973 times and still had a shallower drawdown than holding.
The full grid: sixteen rules on SPY
The grid runs on SPY from November 1993, the first day it has a 200-day average, to 16 October 2026. Buy and hold starts on that day too, so every row covers the same period. Return a year is the compound annual growth rate (CAGR). Worst drawdown is the largest fall from a previous peak to a later low.
CAGR over annualised volatility divides return a year by the size of daily swings. A change of exposure is one switch between the fund and cash.
Source: YX Insights
Chart 2 plots buy and hold and the ten pairs. Buy and hold sits furthest left, with the highest return. Every pair sits below it and to its right.
Rule | Return a year | CAGR over annualised volatility | Worst drawdown | Changes of exposure |
|---|---|---|---|---|
Buy and hold | 10.84% | 0.58 | -55.2% | 1 |
10-day | 0.64% | 0.05 | -64.7% | 1,464 |
20-day | 2.66% | 0.24 | -49.6% | 973 |
50-day | 4.03% | 0.36 | -47.0% | 583 |
100-day | 6.37% | 0.55 | -52.7% | 359 |
200-day | 8.39% | 0.69 | -28.9% | 211 |
10/20 | 4.49% | 0.39 | -37.8% | 409 |
10/50 | 6.25% | 0.54 | -28.9% | 205 |
10/100 | 8.40% | 0.70 | -33.2% | 113 |
10/200 | 9.76% | 0.78 | -20.3% | 61 |
20/50 | 5.43% | 0.45 | -39.5% | 171 |
20/100 | 6.65% | 0.54 | -37.1% | 95 |
20/200 | 9.37% | 0.72 | -33.3% | 49 |
50/100 | 7.57% | 0.58 | -34.3% | 67 |
50/200 | 9.93% | 0.72 | -33.7% | 31 |
100/200 | 9.44% | 0.66 | -33.7% | 27 |
Pairs are listed by fast, then slow length. The order is not a ranking. The CAGR-over-volatility column is rounded to two decimals, so 50/100 and buy and hold both read 0.58.
On SPY, each of the ten pairs earned more than either single length it is built from. Each also changed exposure fewer times than either. A pair switches only when two averages cross. The 10/200 pair changed exposure 61 times in 32.9 years.
The pair with the highest return is not the pair with the shallowest drawdown. Across the seven funds, the pair with the highest CAGR over annualised volatility was one of six different pairs.
The rules that beat buy and hold on a full history
On full histories, seven of the 105 rules beat buy and hold on return, six of them pairs, across three funds. Not one beats buy and hold in a window that holds no crash. It is a different crash on each fund.
QQQ, three rules. Buy and hold on QQQ suffered a drawdown of 83.0% in the dot-com crash. Avoiding that one episode is the whole edge. All three lose to buy and hold from every later start date we tested.
TLT, three rules. Their lead comes from the 2020 to 2023 bond rout. Remove the rout from the window and all three trail buy and hold.
SLV, one rule. The whole edge is the 2011 silver spike, so we count it as a result of that spike. From the 2012 start date its lead is 1.10 percentage points a year.
In each case the lead comes from one fall in that fund's history. It says little about the rule.
How the trend model in Multi-model Signals uses this
Multi-model Signals is our published set of trading signals. One of its models follows trends. It does not apply one pair to every fund, because the choice is made asset by asset. The rules tested here are not the ones behind any signal we publish.
What this test can and cannot show
Cash earns nothing. A rule out of the market would have earned something on that cash, so every rule is understated. Crediting cash with Treasury bill returns from 2007 moved the tests beating holding on return from 7 to 11 of 105, so the finding survives.
The CAGR-over-volatility column is not a Sharpe ratio. It divides a compounded return by a measure built from daily moves.
Costs. There are no wider spreads in a crash and no tax. Real results for the fastest rules would be worse, strengthening the conclusion.
The funds are not independent. SPY, QQQ, DIA and IWM are four views of one equity market. Gold and silver move together.
The sample is small. The seven funds are large survivors picked in 2026. The results illustrate only. The moving average guide treats cash and costs differently, so its figures differ.
In these tests, most rules gave up return and cut the worst drawdown. The cut varied with the pair, the fund and the window.
Learn more with YX Insights
This explainer is part of the YX Insights Academy. Each one takes a single idea and checks it against real data.
The same approach runs through everything else we publish:
Systematic Portfolio: ready-made portfolios for Macro & Megacaps and for Commodities. We publish the holdings and every change.
Multi-model Signals: a daily long-or-flat call on every name we cover. Each call shows how strongly our four models agree.
Research: company deep dives, macro commentary and essays on how we test.
Good places to start on the website:
Academy: more explainers like this one.
Start here: get our free weekly letter by email.
Macro & Megacaps Systematic Portfolio and Commodities: the two portfolios in detail.
Track record: live results since 3 August 2026, plus the ten-year backtest.
Upgrade: get the Systematic Portfolio, Multi-model Signals and research in full.
Common questions about moving average crossovers
Which moving average crossover is best?
No pair came out best on every measure or every fund. On SPY, the pair with the highest return was not the pair with the shallowest drawdown.
Does the golden cross work?
Not on return. The golden cross is the 50/200 pair. It trailed buy and hold on six of the seven funds, leading only on QQQ. What Is a Golden Cross? covers it.
Do moving average crossovers reduce drawdowns?
In these tests, mostly yes. Between 82 and 100 of 105 rules had a shallower worst drawdown than buy and hold, depending on the start date.
DISCLAIMER: This article is strictly educational. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.