Chart patterns are everywhere: head and shoulders, double tops, triangles, flags, cup and handle. Books, courses and social feeds show them with confident arrows pointing to the next move. The question that matters is simpler and harder: does a given pattern tell you where the price goes next, well enough to make money after costs? This guide explains why the answer is less clear than the marketing suggests, and how you can test a pattern yourself without fooling yourself.
Why patterns feel like they work
Patterns feel reliable for several reasons, and each one is a trap.
The pattern is drawn after the move. Once a price has fallen, a chart can be marked with a head and shoulders that now looks obvious. Before the fall, the same chart was ambiguous, and a trader could have drawn three different patterns on it.
Only the successes are remembered. A pattern that works once is shared. A pattern that fails ten times is quietly forgotten.
The pattern is always "confirmed" in hindsight. Each example in a textbook shows a clean pattern followed by a clean move. Real charts are messy, and the clean examples were chosen to make the point.
Random charts also produce patterns. Price series wander up and down, and any long wandering series will contain shapes that look like famous patterns. The eye finds them easily, even when nothing predictable is happening.
These traps do not mean patterns are useless. They mean a pattern's reputation tells you very little about its record.
What the research suggests
Researchers have tested chart patterns on past price data for decades. The results are mixed, and the way a test is run matters a great deal.
Some studies have found that certain patterns carry a small amount of predictive information in some markets over some periods. Others have found that the effect is too small to trade once costs are included, or that it varies from one period to the next. Results that look good on paper often shrink when the rules are fixed in advance and applied to data the pattern was not chosen from.
The fair summary is this: patterns are not proven to predict prices reliably, and claims that they do should be treated with caution until someone shows a test that holds up out of sample, after costs, and over a long period.
How to test a pattern fairly
If you want to know whether a pattern helps, the process matters more than the result. A fair test has these features.
- Write the rules before looking at the data. Define the shape precisely: the number of peaks, the minimum height, the tolerance for the neckline, and the exact entry and exit points.
- Use data you did not use to design the rules. If you looked at a period and then wrote the rules, the results on that period are contaminated. Keep a period in reserve.
- Include costs. Add the commission, the spread and any tax to every trade.
- Count every signal. Do not skip the trades that lost. Do not add only the good cases after the fact.
- Compare with a simple alternative. A pattern should beat a plain rule, such as holding, or buying after a rise and selling after a fall. If it does not, the pattern is adding noise.
- Check the sample size. Ten trades tell you almost nothing. Hundreds of trades across different conditions tell you more.
A worked example of a fair test
Suppose you want to test a simple rule: buy when the price closes above the high of the last 20 days, and sell after 10 days. You take five years of daily prices for one stock. You write the rule down, then run it on the first four years and check the result on the last year, which you did not use to design the rule.
Imagine the rule makes 120 trades over the test period. Each trade has a small expected gain before costs. After commissions and the spread, the average gain per trade is close to zero. Over 120 trades the total result is slightly negative, and the same rule on a random walk, generated with the same volatility, produces a similar result.
That outcome is not a failure of your analysis. It is the answer. The rule has not beaten chance after costs, so you should not trade it on the expectation that it will work.
| Test step | What to record | Why it matters |
|---|---|---|
| Rule | Exact entry and exit | Removes hindsight |
| Data split | Design period and test period | Tests on data the rule did not see |
| Costs | Commission and spread per trade | Shows the net result |
| Count | Every signal, win or loss | Avoids cherry-picking |
| Benchmark | A simple rule or a random walk | Shows whether the pattern adds anything |
Market conditions matter
A pattern can behave differently in a strong trend than in a sideways market. A continuation pattern may do well when the trend is strong and poorly when the price is stuck in a range. Testing a pattern across all conditions together can hide that. Split the data by condition, such as trending and ranging periods, and check whether the result holds in each.
The same applies to volatility. A pattern that works in quiet markets may fail in fast ones, when the price gaps through the levels the pattern depends on. A result that looks strong over ten years may be made of two good years and eight poor ones. Look at the spread of results across periods, not just the average.
What this means for a trader
Patterns can still be useful as a way to structure a plan. A pattern can give you a clear entry, a place to put a stop and a way to size a position. Those are valuable, even if the pattern does not predict the next move. A trader with a clear plan and strict risk control can do well with a pattern that has no predictive edge, and a trader without that discipline can lose money with the best pattern in the world.
The distinction matters. A pattern is a tool for managing a trade. Whether it predicts prices is a separate question, and the answer should come from a fair test, not from the examples you remember.
Common mistakes
Testing on the data you used to design the rule. This is the most common error, and it makes almost any idea look good.
Ignoring costs. A pattern that gains a little before costs can lose after them. Always include costs.
Stopping the test when it looks good. Decide the test length and the number of trades before you start. Stopping early because the result looks good is another form of cherry-picking.
Confusing a pattern with a signal. Seeing a pattern is not the same as the price responding to it. Measure what the price does after the pattern, against what it does at other times.
Key takeaways
Chart patterns are popular, and their reputation is stronger than their record. Research on them is mixed, and the examples most people see were chosen after the fact. To judge a pattern, write the rule first, test it on data you did not design it from, include every cost, count every signal and compare the result with a simple benchmark. If a pattern does not beat chance after costs, it may still help you manage trades, but it should not be trusted as a forecast.
This guide is for general information, not financial advice. See the disclaimer.