← All lessons

A pattern on trial

Turn “hammers work” into a claim that data can confirm or reject. · 7 min

Every chart book says the hammer, a candle with a long lower wick after a decline, marks a bottom. Here are ten years of daily candles from our simulated market. Every hammer is marked, and the five days after it are shaded.

80.7870.7260.65+1.9%-3.9%+3.3%+1.0%
Hammer signals (triangles), the 5 days after each (shaded) and the return from the next open to the close 5 days later.

Simulated data from the course’s market engine, not real prices.

Predict first

Over the whole ten years, the market printed 49 hammers (counting each once). Five days later, was price higher after more than half of them?

A claim you can test

“Hammers work” cannot be wrong, so it cannot be tested. A testable claim pins down five things before looking at any result:

  1. The signal, as numbers. Ours: the lower wick at least 2 times the body, the upper wick at most 25% of the range, a body larger than a doji’s, and at least 5 lower closes in the 8 candles before it.
  2. When it is known: at the close of the hammer, never earlier.
  3. The trade: buy at the next open.
  4. The horizon: sell at the close five days later.
  5. The comparison: what a random day in the same data earned over the same five days.

Thomas Bulkowski built his encyclopedia of candlestick patterns this way: thousands of patterns, each with its rule, its sample size and what followed. This track gives you the same tools, on data where we know the truth.

Simulated data from the course’s market engine, not real prices. The research market has fat tails, volatility clusters and gaps like a real stock, but no drift and no memory: by construction, nothing predicts its next move. That makes it the perfect courtroom. Any pattern that “works” here is working by luck.

Worked example

The first hammer in the data is on day 131: open 113.24, high 113.24, low 111.89, close 112.88.

  • Range = 113.24 − 111.89 = 1.35; body = |112.88 − 113.24| = 0.36
  • Lower wick = 112.88 − 111.89 = 0.99, which is 2.7 bodies: at least 2, so the shape qualifies.
  • Upper wick = 113.24 − 113.24 = 0.00: short enough. The candles before it were falling, so it is a hammer, not a hanging man.
  • Trade: buy at the next open, 112.88; sell at the close five days later, 112.94. Return = 112.94 ÷ 112.88 − 1 = +0.05%.

Put a pattern on trial

Pick a rule. The lab finds every signal in ten years of daily candles and counts how often price was higher five days later.

Running the numbers…

Common mistake

Judging a pattern by the examples you remember. Books and videos show the hammers that turned into big rallies, because those make good pictures. The ones that failed are not shown. Only a rule applied to every candle, with every result counted, tells you what a pattern does.

Which of these is a testable claim?
A hammer forms on today’s daily candle. When can a test act on it?

Closing round

Edge or noise?

A fresh market and a pattern, with or without a hidden edge planted after each signal. Read the study and call it.

Building a fresh case…

Further reading: Bulkowski, Encyclopedia of Candlestick Charts; Aronson, Evidence-Based Technical Analysis, chapter 1.