Death Cross

What a death cross is

A moving average is the average of the last N closing prices, recalculated every bar. A simple moving average (SMA) weights every bar in the window equally. Because it drops the oldest price and adds the newest each bar, it moves slowly and smooths out the noise.

A short average reacts faster than a long one. In a steady rise the short average sits above the long one, because recent prices are higher than older ones. When price turns down for long enough, the short average falls through the long one. That moment is the death cross. The common pair is the 50-day and 200-day SMA on a daily chart, but the same idea is used with other lengths, with exponential averages, and on other timeframes. See EMA crossover for the version that weights recent bars more.

The name is a label, nothing more. It has no special force. The rule is simply: the fast average was at or above the slow average on the previous bar, and it is below on this bar. The opposite event, the fast average rising above the slow one, is the golden cross. Read that page too, because every limit described here applies to it in reverse.

Why it is late by design

An average of the last 50 closes includes prices from nearly ten weeks ago on a daily chart. The 200-day average reaches back about 40 weeks. For the 50 to fall below the 200, the recent prices must be low enough for long enough to drag the short average under a long one that is itself sitting at higher, older levels. That takes time.

This delay is called lag. A longer window means smoother lines and fewer false alarms, but more lag. A shorter window means less lag but more crosses that reverse within days. There is no setting that removes both problems; each choice moves the cost somewhere else.

So a death cross usually shows up after a fall has already happened. Sometimes the fall is over by the time the lines cross, and the price has started to recover. The cross is then reporting the past. This is the same trade-off every trend-following rule has, and it is why the signal is better thought of as a description of what has happened than a forecast.

A worked example with made-up numbers

Take 300 made-up daily closes. They rise from 100 to a high, then slide lower, with a small wobble on top. We calculate a 50-day SMA and a 200-day SMA of the closes. The 200-day average needs 200 closes, so the first value is on day 200. The table shows four days picked from that run.

The highest close is 162.42 on day 219. The 50-day SMA first falls below the 200-day SMA on day 277, 58 days later. On the day before, the 50 was 141.51 and the 200 was 141.11, so the 50 was still above. On day 277 the 50 is 140.85 and the 200 is 141.14, so it has crossed.

The close on the cross day is 125.34. That is 37.08 points, or 22.8 percent, below the high. So the cross did not mark the top. It arrived after more than a fifth of the price had already gone. After the cross the close drifted down further to 109.36 on day 300, a further 15.98 points (12.7 percent of the close at the cross). From the high, the total fall by day 300 is 32.7 percent. All of these figures are made up and were calculated by us on 31 Oct 2026, exact. One made-up path proves nothing about real markets; it shows what lag looks like in numbers.

How a bot uses a death cross

A bot needs a rule written without any room for judgment. A simple one looks like this, and it is an example of a rule, not a recommendation:

• Condition. On the close of each daily bar, compute the 50-day and 200-day SMA. If the 50 was at or above the 200 on the previous bar and is below it now, the event has happened.
• Action. The bot does one fixed thing on the event: for example it closes any long position, or it opens a short one. Which action, and whether it acts on the same bar or the next, is a setting you must define.
• Re-arm. The bot waits for the opposite event, a golden cross, or for a fixed exit rule, before acting again.

A rule that acts at the close of the bar that made the cross trades at a price it only knew at the end of that bar. A test that uses the cross to trade at that bar's open or at an earlier price is using information from the future. See look-ahead bias.

How it breaks a bot

• Whipsaws. When the two averages run close together, small moves in price can push the fast line across the slow one and back. In a sideways market that gives a string of crosses, each followed by a reversal. In the made-up path above there is 1 cross after day 200, because the move is steady. A real choppy period can produce many more, and each one costs a spread and possibly slippage. A close-based filter such as the one in false breakout is one way people test fewer, later signals, at the price of a later entry.
• Late exits. If the bot only reacts at the cross, it gives up the early part of any fall, as in the example where 22.8 percent was already gone.
• Fees on rare trades. A daily 50/200 rule trades seldom. A few trades a year means a very small sample, and a small sample makes any average result unreliable.
• Settings picked from the past. Trying 20/100, 30/150, 50/200 and keeping the best on old data is overfitting. The chosen pair may only describe the history it was fitted to.

How to check it properly

Write the rule in full, including the action and the bar on which it acts. Run it on years of real data across rising, falling and flat periods, with spread, fees and slippage included. Record how many crosses happened, how the position performed after each, and the worst stretch from peak to trough. Compare the result with simply holding, and with the same rule using another pair of lengths, using expectancy rather than a single total. Then test on a stretch the rule has never seen, as in forward testing. The full method is in how to backtest.

Because a death cross is slow, a bot using it usually sits beside rules that act faster, such as a trailing stop, and beside sizing rules from the position size calculator. Other tools describe different things: RSI measures the pace of recent moves, Bollinger Bands measure spread around an average, and VWAP is an average weighted by volume over a session. None of them is a better version of a moving average cross; they answer different questions.

What a death cross does not tell you

It does not say that price will keep falling, that it will recover, or how far either will go. It does not say how often it has been followed by more losses, because that varies with the market, the period and the lengths chosen. It describes the relationship between two averages of old prices, nothing else. This page is education only.