# Do Candlestick Patterns and Double Bottoms Work — Testing 5 Textbook Patterns
- Source: FoldAlpha Research (https://app.foldalpha.com/en/research/candlestick-patterns) · Published: 2026-09-01 · Series: Myth Testing 7

Candlestick reversal patterns and double bottoms are staples of chart-education content. We tested them exhaustively against historical data.

## Test 1 — Four Textbook Candlestick Patterns

Shooting star, dark cloud cover, bearish engulfing, and long bearish candle —
we detected them mechanically according to their textbook definitions and compared post-pattern returns against days with no pattern (the baseline).

| Pattern | Sample | Verdict |
|---|---|---|
| 4 candlestick reversal patterns | 15–133 occurrences each | **No statistical difference in subsequent returns from the baseline** |

We note plainly that the samples themselves are not large. That said, all four pointed in the same direction (no difference).

## Test 2 — KOSPI Double-Bottom Neckline Breakouts

We defined double bottoms mechanically on the index chart (conditions on the spacing and depth between the two lows), then bought on the day of the upward neckline breakout:

| | Sample | 60-day win rate |
|---|---|---|
| Buy after double-bottom neckline breakout | 42 occurrences | 55% |
| Buy at a random point in the same period (baseline) | — | **59%** |

Buying the breakout was **actually lower** than buying on any random day. The 60-day return was inferior as well.

## What These Numbers Say

Candlestick patterns are widely discussed not for their predictive power but for their **after-the-fact explanatory power**. Look at a chart after a decline and a reversal pattern is always "there" — because patterns are everywhere. To use them for prediction, forward returns would have to differ at the moment the pattern appears, and no such difference was observed.

The double bottom's 55% is worth noting as well. The index rises on more days than not to begin with (baseline 59%), so "this signal is right more than half the time" reflects a basic property of the market, not an effect of the signal. **Win rates must always be compared against the baseline.**

## Limitations

- The pattern definitions are mechanical implementations of textbook criteria; discretionary chart interpretation ("an eye for context") cannot be tested — and being untestable is itself a characteristic of such claims
- The candlestick pattern samples (15–133 occurrences) are small. The result is closer to "if there were an effect, it would not look like this" than to "there is an effect and we failed to catch it"

## Data Sources

- Korea Exchange daily prices (index and individual stocks). Long-term index trends can be viewed on the [market page](/market)

---
This article documents tests on historical data for informational purposes only. It is not investment advice or a recommendation to buy or sell any security. Past test results do not guarantee future returns.