What Is a Death Cross?

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-15 · Updated 2026-10-03 · Reviewed 2026-09-26

A death cross is the day a price’s 50-day moving average closes below its 200-day moving average, a signal usually read as bearish. On the S&P 500 it has printed 38 times since 1950, and at the median cross the index was already 8.4% under its highest close of the prior year.

It gets headlines because of what it is called. What it actually reports is a decline that has already been going on for months, and a year later the index had done about as well after a cross as after any other day.

How it works

Two averages are drawn on a daily chart. One covers the last 50 closes, the other the last 200, and both are recalculated as each new day arrives.

The 50 responds faster. A new close is a fiftieth of the short average and a two-hundredth of the long one, so recent weakness moves it much further.

The cross happens once enough closes have fallen. For a 50-day average to sit below a 200-day one, roughly the last quarter has to have been meaningfully weaker than the last year.

Why the timing is the whole story

The signal is a summary of the last few months. It cannot appear early, because the averages only move once the closes that would move them have printed.

Which means it arrives at a point that could be anywhere in a decline. Sometimes near the start of a long one, sometimes near the bottom of a short one - and the cross itself carries no marker telling you which.

The headline treats it as a warning. It is a description, and a description of a period that has already closed.

Is a death cross bearish?

For the next few months, mildly, on the S&P 500’s record since 1950. Since then the index rose less often in the one to six months after a cross than after an ordinary day, which is the direction the name suggests.

Over a full year, the difference disappears. The declines that produce a cross are often well along, or already over, by the time the two lines meet, so the recovery is frequently inside the next twelve months.

So it marks a weak stretch that has happened, not one that is coming. Whether the next few months are worse depends on which kind of decline you are in, and the cross looks identical in both. The exact figures, and how small the sample is, are in the data section below.

A worked example

Imagine a market that falls for six weeks and then stabilizes. The 50-day average drops through that period while the 200-day barely moves.

The cross may print weeks after the fall stopped. By then the recent closes are flat, but the average still carries the weak ones, so the short line keeps descending toward the long one.

Selling at that cross sells after the decline. Not because the signal was wrong - it correctly reported that the last quarter was weak - but because reporting the last quarter is all it can do.

The reverse happens too. In a decline that continues for a year, the same cross arrives early enough to be useful, and nothing distinguishes the two cases at the moment the lines touch.

Two real crosses on the S&P 500

April 2025 was the late kind. The index closed at a record 6,144.15 on 19 February 2025 and at 4,982.77 on 8 April, 18.9% lower. The cross printed on 14 April, when the 50-day average (5,748.25) finished under the 200-day (5,753.90) with the index at 5,405.97, four sessions after the lowest close. The lowest close in the following year was 5,158.20 on 21 April, 4.6% under the cross, and the index set a new closing record on 27 June 2025.

March 2022 was the early kind. The cross printed on 14 March 2022 with the index at 4,173.11, 13.0% under its 3 January record of 4,796.56. The slide continued to 3,577.03 on 12 October, another 14.3%, and a year after the cross the index was still 6.7% lower.

On the day, the two looked alike. Each came with the index 12% to 13% off its high and the 50-day just under the 200-day. Nothing in the lines separated the cross near a bottom from the cross near the start of a bear market.

The original data

The test. Daily closes of the S&P 500 index, price only, from 3 January 1950 to 25 September 2026, downloaded from Yahoo Finance. Both lines are simple averages of closes, and a death cross is the first close where the 50-day sits below the 200-day after sitting at or above it the day before. The first checkable day is 19 October 1950, which leaves 75.9 years and 38 crosses, about one every two years. Each one is listed in the death cross table.

Most of the fall came first. At the median cross the index was 8.4% below its highest close of the previous 252 trading days. 34 of the 38 came at least 5% below that high, and 13 at least 10% below.

Weaker for six months, ordinary at a year. “Any day” is every close from 19 October 1950 that has the full period after it, about 19,000 of them. Up to six months out, the index rose less often after a cross than after an average day. At twelve months the gap had closed: 71.1% against 74.4%, and a median change of +10.54% against +10.44%.

Thirty-eight is a small sample, so read the gaps with care. A two-sided binomial test of each share against the any-day rate gives p = 0.094 at one month, 0.30 at three, 0.050 at six and 0.71 at twelve. Only the six-month gap sits near the usual 5% line for chance, and with four periods checked, one result that close is weak evidence.

What followed in the next year. The lowest close in the 252 trading days after a cross was a median 8.5% below the close on the cross day. 15 of the 38 were followed by a further fall of 10% or more and 4 by a further 20% or more, while 11 saw less than 5% of extra downside. And 13 of the 38 were undone by a golden cross within 63 trading days, roughly three months.

Every big decline had one, and so did plenty of smaller ones. Counting each decline from a closing high (the highest close since 1950 at the time) to the first close back above it, all 11 declines of 20% or more that began after October 1950 contained a death cross between the peak and the recovery, and so did 22 of the 25 declines of 10% or more. But 16 of the 38 crosses came outside any 20% decline, and in 5 of those 22 declines of 10% or more, the first cross printed after the lowest close.

The S&P 500’s worst decline since 1950 shows the scale this signal is built for. The index fell 56.8% from its close on 9 October 2007 to its close on 9 March 2009, a slide of about seventeen months — long enough for a 50-day average to fall through a 200-day one somewhere along the way.

Most declines are far shorter than that. Between 1950 and 2026 the index had 68 separate stretches of closing 5% or more below its record, about 0.89 a year, and a short, shallow one can be over before the slower average has moved much at all - which is exactly the case where the cross prints late.

The cost side is unusual here. A trader acting on this signal trades rarely, so the round trip is small beside the position - say 0.1% of it, once or twice a year. The risk is not overtrading, it is acting decisively on a reading whose timing relative to the decline is unknown.

The variations nobody agrees on

The 50 and the 200 are conventions, not a standard. Some versions use simple averages, others exponential, and the two cross on different days from the same data.

The timeframe is also a choice. The named version is a daily chart, but the same crossover exists on weekly bars, hourly bars and any other interval, and each produces a different set of dates.

Which means two people can disagree about whether it has happened. One runs exponential averages and sees the cross on Tuesday; the other runs simple ones and is still waiting.

No arbiter settles it. There is no body that defines the pattern, so the version that gets reported is whichever one a given outlet happens to compute - a detail rarely mentioned alongside the headline.

Why the name matters more than it should

The label is doing the persuading. “Death cross” is not a technical term with a definition beyond the 50 crossing the 200 - the two words attached to it are editorial.

Frightening names travel. A crossover with a dramatic name gets covered by financial media, which puts it in front of people who are not watching the chart, which is why it is one of the few technical events with a general audience.

That coverage is its own effect. If enough people act on a widely reported signal, the signal moves price for reasons unrelated to the averages that produced it.

Which is worth separating from the original claim. “This pattern predicts declines” and “this pattern gets reported so widely that people react to it” are different statements, and only the second one has an obvious mechanism behind it.

How this connects

The two lines are usually a simple moving average, although some charting packages use an exponential moving average instead, which reacts faster and crosses on different days.

The slow line on its own has a page too. The 200-day moving average is the level many investors watch even without a second line to cross it.

It is the classic signal of a trend following system. Those systems accept late entries and late exits by design, trading timeliness for the chance of staying on the right side of a long move.

The cross tends to appear inside a bear market or a market correction, but it cannot tell the two apart, and one lasts far longer than the other.

After a sharp fall, a dead cat bounce can lift the short average for a while and bring the lines close together again without the decline being over.

Anyone tempted to act on it should backtest the rule first and be wary of overfitting, since tuning the two lengths to past crashes produces a rule that describes history rather than one that holds up on new data.

It is also a blunt tool for market timing. A buy and hold investor who sells on the cross has to be right twice, once on the way out and again on the way back in.

And confluence is the usual answer to its lateness, pairing the cross with price structure such as a lower swing or a broken support and resistance level.

When it fails

The characteristic failure is selling the bottom. The averages need months of weak closes to cross, so the signal often prints after the sharpest part of a decline is over.

The seller then watches the recovery from outside. The cross was an accurate description of the previous quarter and a poor guide to the next one, and the difference between those two only becomes visible afterwards.

A second failure is treating it as rare and therefore meaningful. It is rare because the periods are long, not because the condition it describes is unusual.

A third is acting on the headline rather than the chart, by which point the cross is days old.

A fourth is applying it to a single stock, where the 50 and 200 were never conventions in the first place.

And a fifth is counting the hits after the fact. Every decline of 20% or more in the record above contains a death cross somewhere, because a long enough fall drags a 50-day average below a 200-day one. That makes the pattern present at every crash and says nothing about how often it appears without one, which is the number to ask for before trusting any list of famous crosses.

Related

Moving average crossover covers the general mechanism. Golden cross covers the upward version with the opposite name. And moving average covers the lines both are built from.

The size of the falls themselves is measured on the drawdown page, and the 200-day moving average page tests rules built on the slower line alone.

This page is educational, not financial advice. Test every idea on your own charts before risking money.