The golden cross is probably the most repeated signal in retail trading: when the 50-day moving average crosses above the 200-day, the chart is said to turn bullish. It has everything a famous signal needs — a name, a clean picture, and a century of folklore. What it rarely comes with is a full track record: every writeup shows the same two or three beautiful crosses and stops there.
So we ran it properly. Below are the exact rules we tested, the live results from our engine, and — more useful than either — where this strategy tends to break.
The exact rules we tested
“Golden cross” means slightly different things in different writeups, so here is our template’s spec, stated precisely. This matters: a backtest you can’t restate in exact rules is a backtest you can’t trust.
- Universe: SPY, QQQ, AAPL, MSFT, NVDA — daily bars.
- Entry:buy on the day the 50-day simple moving average closes above the 200-day, having been at or below it the day before (the cross day itself, not just “50 above 200”).
- Exit:sell when the 50-day closes back below the 200-day — the “death cross.” No stop-loss, no profit target: this is the pure signal, unedited.
- Sizing: 20% of equity per position, up to 5 positions at once.
What the numbers mean
The first thing to notice is how few trades a decade produces. Moving averages this slow cross rarely — this is a strategy that acts a handful of times per symbol per decade and holds for months or years. That has two consequences worth sitting with.
First, the sample is small. A strategy with this few round trips can look brilliant or broken because of two or three market regimes, not because of any deep statistical edge. Judge it as a trend filter with a track record, not a machine with an expected value per trade.
Second, the golden cross is structurally late twice. The 50/200 cross confirms an uptrend only after months of rising prices, so you buy well off the low; the death cross confirms the downtrend the same way, so you sell well off the high. What you get in exchange is time in the big moves and time out of the big disasters. Whether that trade-off pays is exactly what the drawdown and CAGR numbers above are telling you.
Where it breaks
The classic failure mode is a sideways market. When price chops in a range, the averages braid around each other and the strategy buys the cross, sells the re-cross lower, and repeats — a sequence of small, demoralizing losses that the folklore never mentions. Long stretches of the record are exactly that.
Variations worth testing
The unedited signal is a baseline, not a finish line. In the lab you can change any rule in plain English and rerun in seconds. The edits people most often try from here:
- “Add an 8% stop-loss” — caps the worst single-name damage.
- “Use 20/100 instead of 50/200” — faster crosses, more trades, more whipsaw; watch the drawdown.
- “Only enter when price is also above the 200-day” — a second confirmation layer.
- “Test 2006–2013 instead” — a window with a very different character (our daily data starts in 2006); regime changes move slow strategies more than fast ones.
And when a variation survives your skepticism, the real test isn’t another backtest — it’s freezing the rules and letting it trade forward on data nobody has seen. That’s what our public ledger is for.