Candlesticks and chart patterns: what the evidence says
Chart patterns and candlesticks are the most recognisable part of technical analysis, and the most debated. This lesson covers what they are, why people use them, and what the research actually finds.
Candlesticks
A candle shows four prices for a period: open, high, low and close. The body spans open to close, the wicks show the extremes. The shape summarises who controlled the session.
| Candle | Shape | The story it tells |
|---|---|---|
| Doji | Open and close almost equal | Indecision: neither side won the session |
| Hammer / shooting star | Small body, long lower / upper wick | A push one way was rejected |
| Engulfing | Body fully covers the previous candle | Control switched sides |
| Marubozu | Large body, little or no wick | One side dominated all session |
Do candles have predictive power?
On their own, mostly no. A widely cited study (Marshall, Young and Rose, 2006) tested classic Japanese candlestick patterns on large US stocks over a decade and found no profitable signal after accounting for chance. Later studies find some effects in specific markets or periods, but they tend to be small, inconsistent and eaten by trading costs.
Where candles do help is context. A long lower wick at a level where large buyers have stepped in before, on heavy volume, says something. The same candle in the middle of nowhere says very little. The candle is a description of what just happened, not a forecast.
Chart patterns
| Pattern | What it looks like | Usually read as |
|---|---|---|
| Bull / bear flag | Sharp move, then a short, tight counter-drift | Pause before the trend continues |
| Cup and handle | Rounded base, then a small dip near the old high | Accumulation before a breakout |
| Head and shoulders | Three peaks, the middle highest | Uptrend losing strength, possible reversal |
| Double top / bottom | Two failed attempts at the same level | Level holding, possible reversal |
| Triangles and wedges | Range narrowing between converging lines | Volatility compressing before a move |
Do patterns work?
The evidence is mixed, and better than for single candles. Lo, Mamaysky and Wang (2000) used an algorithm to detect patterns like head and shoulders objectively, and found several did carry some information about future returns, especially in smaller Nasdaq stocks. But the effects were modest, and practitioner statistics on pattern “success rates” are usually descriptive, not tested out of sample.
Three problems hold patterns back:
- Subjectivity: is that a flag or just noise? Two people often disagree, and patterns are far clearer in hindsight.
- Survivorship: the failed patterns are forgotten, the textbook ones get posted.
- Crowding: when everyone watches the same breakout level, stop orders cluster there, which can cause false breakouts as easily as real ones.
How to use them sensibly
- As risk frameworks: a flag gives a clear invalidation point (the bottom of the flag) and a clear trigger (the breakout).
- With trend: continuation patterns in the direction of a strong trend hold up better than reversal calls against it, consistent with momentum (lesson 5.6).
- With volume: breakouts on rising volume are more credible (lesson 5.1).
- With a reason: a pattern forming into earnings, a macro event or a known positioning level means more than one in isolation.
Key takeaway
Candles describe, they don't predict. Patterns can offer a slight edge and, more usefully, a clear plan for where you are wrong. Treat them as tools for structuring trades, not as signals on their own.
Educational content only, not financial advice. Investing involves risk and you can lose money.