Study · October 2026
Does the golden cross work? We tested it
Buy when the 50-day average crosses above the 200-day one, sell when it falls back below. It is the most famous trend rule there is. On years it never saw, its trades made money. And it still did worse than simply holding the same stocks.
How we tested it
We wrote the test down before downloading a single price: the rule, the stocks, the years and what would count as working. Three textbook pairs were allowed (50/200 simple, 20/50 and 10/30 exponential). The pair with the best result on the Dow 30 from 2016 to 2021 would be chosen, and only then measured, once, where it had never been looked at: the same stocks from 2022, and 20 other large caps from 2016.
Each trade had a stop three average daily ranges below the entry, and paid a fee and slippage on each side. Results are in R: the amount risked on a trade. +1 R means the trade made what it risked.
The trades looked good
The 50/200 pair won the choice. On the Dow 30 from 2022, over 116 trades, it made +1.75 R per trade on average after costs (with a 95% interval of ±1.25 R), even though only 29% of trades won: the winners ran far and the losers were cut. On the 20 held-back stocks, 141 trades made +2.5 R per trade.
Holding the stock did better
Per-trade profit is not the whole question. The rule sits in cash much of the time (it was invested on 44% of days), so we compared each stock's rule curve with simply holding that stock over the same days.
| Golden cross | Holding | |
|---|---|---|
| Dow 30, 2022–2026: median Sharpe ratio | 0.37 | 0.65 |
| Median annual return | +4.8% | +12.7% |
| Median worst fall | −25% | −35% |
| Stocks where it did at least as well as holding | 7 of 30 | |
| 20 other large caps, 2016–2026: median Sharpe ratio | 0.41 | 0.60 |
| Stocks where it did at least as well as holding | 2 of 20 | |
The test asked for at least half the stocks. It managed 7 of 30, then 2 of 20. Verdict: does not hold up. What the rule does give is smaller falls, which is a real benefit, but a different goal from earning more than holding.
Could a tweak save it?
We tried two popular ones afterwards, knowing that anything chosen after seeing results needs fresh data to count. Trading only "clean" trends (a trend-strength filter) made it worse: on the Dow 30 from 2022 its median Sharpe ratio fell below zero. "Never sell below the entry price" left positions open for years at deep losses, one of them down 75%. Neither fixed the problem.
What this means for your rules
- A rule can make money per trade and still be worse than doing nothing clever at all.
- A low win rate is not a flaw, and a high one is not proof: what counts is the whole curve against an honest benchmark.
- Both stock lists are today's large companies, which flatters any rule that buys them and holding alike. Past prices; not a forecast and not advice.