Learn the stock market › Is it true?
Does technical analysis work? We tested five well-known signals
On the 30 largest Stockholm companies since 2000, almost nothing could be told apart from an ordinary day: 1 of 20 comparisons fell outside the range of chance, about as many as chance gives (RSI over 70 after one week, −0.3 points). Three months after a death cross the stock had risen in 63.6% of cases, after a golden cross in 63.8%. On an ordinary day: 62.0%.
Difference from an ordinary day, one week later
Difference from an ordinary day, one month later
Difference from an ordinary day, three months later
Difference from an ordinary day, twelve months later
| Signal | Times | Rose after one week | Ordinary day | Average | Ordinary day |
|---|---|---|---|---|---|
| Golden cross | 441 | 51.5% | 53.2% | +0.2% | +0.3% |
| Death cross | 435 | 54.0% | 53.2% | +0.5% | +0.3% |
| RSI under 30 | 973 | 57.8% | 53.3% | +0.8% | +0.3% |
| RSI over 70 | 1,834 | 49.2% | 53.4% | +0.0% | +0.3% |
| MACD cross | 4,532 | 53.2% | 53.3% | +0.3% | +0.3% |
| Signal | Times | Rose after one month | Ordinary day | Average | Ordinary day |
|---|---|---|---|---|---|
| Golden cross | 441 | 52.2% | 57.1% | +0.7% | +1.3% |
| Death cross | 435 | 58.9% | 57.0% | +1.9% | +1.3% |
| RSI under 30 | 967 | 60.9% | 57.2% | +1.9% | +1.3% |
| RSI over 70 | 1,829 | 55.5% | 57.5% | +0.9% | +1.4% |
| MACD cross | 4,517 | 56.5% | 57.4% | +1.1% | +1.3% |
| Signal | Times | Rose after three months | Ordinary day | Average | Ordinary day |
|---|---|---|---|---|---|
| Golden cross | 434 | 63.8% | 62.0% | +4.0% | +4.0% |
| Death cross | 431 | 63.6% | 62.0% | +4.5% | +4.0% |
| RSI under 30 | 962 | 63.9% | 62.1% | +4.8% | +4.0% |
| RSI over 70 | 1,818 | 62.5% | 62.7% | +3.6% | +4.2% |
| MACD cross | 4,487 | 61.7% | 62.5% | +3.8% | +4.0% |
| Signal | Times | Rose after twelve months | Ordinary day | Average | Ordinary day |
|---|---|---|---|---|---|
| Golden cross | 423 | 68.6% | 69.9% | +16.4% | +17.2% |
| Death cross | 419 | 69.5% | 69.9% | +16.2% | +17.3% |
| RSI under 30 | 934 | 69.3% | 70.0% | +19.7% | +17.4% |
| RSI over 70 | 1,743 | 72.1% | 70.8% | +17.3% | +18.2% |
| MACD cross | 4,329 | 70.7% | 70.5% | +17.6% | +17.5% |
What the signals are said to mean, and what happened
These numbers are for three months after the signal.
Golden cross
The 50-day average moves above the 200-day (how the crosses work). Said to be a sign that a rise has begun. It happened 434 times. Three months later the stock had risen in 63.8% of cases, against 62.0% on an ordinary day. On average +4.0% against +4.0%.
Death cross
The 50-day average moves below the 200-day. Said to be a sign that a fall has begun. It happened 431 times. Three months later the stock had risen in 63.6% of cases, against 62.0% on an ordinary day. On average +4.5% against +4.0%.
RSI under 30
The stock has fallen fast and is called oversold. Said to be a sign that it will bounce. It happened 962 times. Three months later the stock had risen in 63.9% of cases, against 62.1% on an ordinary day. On average +4.8% against +4.0%.
RSI over 70
The stock has risen fast and is called overbought. Said to be a sign that it will fall back. It happened 1,818 times. Three months later the stock had risen in 62.5% of cases, against 62.7% on an ordinary day. On average +3.6% against +4.2%.
MACD cross
The fast MACD line moves above its signal line. Said to be a sign that upward momentum is growing. It happened 4,487 times. Three months later the stock had risen in 61.7% of cases, against 62.5% on an ordinary day. On average +3.8% against +4.0%.
But these companies are winners
The 30 companies are the largest in Stockholm today. Companies that fell and never came back are missing, and that flatters signals that come after a fall. So we ran the same test on the S&P 500 since 1928, where nothing has been sorted out.
There it leans the way the signals promise: after a golden cross the index rose in 70.0% of cases, after a death cross in 54.0% and on an ordinary day in 63.9%. But that is only 50 times each.
For both crosses the range includes zero, so it cannot be told apart from chance.
The average after a death cross was also higher than on an ordinary day, +3.2% against +2.0%, because the times the market rose afterwards it rose a lot.
Across all holding times, 1 of 20 comparisons on the S&P 500 fell outside the range of chance.
| S&P 500, 3 months | Times | Rose | Ordinary day | Average difference |
|---|---|---|---|---|
| Golden cross | 50 | 70.0% | 63.9% | +1.8% |
| Death cross | 50 | 54.0% | 63.9% | +1.2% |
| RSI under 30 | 178 | 60.1% | 63.9% | +0.1% |
| RSI over 70 | 297 | 69.0% | 63.9% | +0.4% |
| MACD cross | 643 | 63.8% | 63.9% | +0.1% |
A week is not three months
We also tested the shortest holding time, one week. What stood out on the Stockholm companies: RSI over 70 after one week, −0.3 points.
The S&P 500 does not show the same thing.
With 20 comparisons you expect about one outside the range from chance alone, so a single one says little. Fees and tax are not counted either.
A thesis from Lund University (Nordström and Andersson, 2008) concluded that RSI, MACD and Bollinger Bands worked on the same kind of companies in 2000–2007. It held positions for five days, shorted on sell signals and compared with buying and holding, not with an ordinary day. Alone, the indicators gave less than buying and holding, 25–54% against 208%. Only just over half of the signals, 51–52%, made money. The combined model beat buying and holding only once the two best stocks were removed, 102% against 78%, and the thesis could not reject that the market is efficient. So the results are not as far apart as the headlines sound. The biggest difference is what you compare with.
What the test does not say
We tested each signal on its own, on large companies and with four holding times. Bollinger Bands are not included. Someone who combines several signals or uses stop levels gets a different outcome. You can test your own rule.
How we counted
- The 30 companies in OMX Stockholm 30 today, from 2000 or from their listing, to Nov 6, 2026. Prices include dividends. The index does not.
- Averages of 50 and 200 days. RSI over 14 days. MACD with 12, 26 and 9 days.
- At most one signal of each kind per company per month (21 trading days), so that one event is not counted several times.
- An ordinary day is the same company on all days in the same period, weighted like the signals.
- The holding times are 5, 21, 63 and 252 trading days.
- The range is calculated on whole years, because many companies signal at the same time. That makes it wider than if each signal counted alone.
- Fees and tax are not included.
Source
- Nordström and Andersson, Position under fem dagar: en vinnande strategi inom teknisk analys?, Lund University, 2008 (in Swedish).
You are welcome to use the numbers
Link here, quote the table or embed the chart on your own page. Name ChartCompass as the source and keep the link.