Padrões de candles: quais têm uma vantagem real e como testar você mesmo
Quick answer: no candlestick pattern works simply because it is that pattern. What sometimes does have an edge is a specific pattern, at a specific level, in a specific session. And you only know by measuring: define the pattern unambiguously, collect 100 cases, and compare the result against what would have happened without the pattern. If it does not beat that baseline, you do not have a signal.
Pontos principais
- The pattern is not the signal; the pattern plus the context is.
- If you cannot code the definition, it is too vague to measure.
- Without a baseline to compare against, every pattern looks like it works.
- 100 cases is the minimum. At 20 you are admiring attractive noise.
What each candle actually describes
| Pattern | What it describes | When it stops meaning anything |
|---|---|---|
| Hammer | Rejection of lower prices within the candle | Mid-range, with no level beneath it |
| Engulfing | One side absorbing the previous range | When the prior candle was tiny |
| Doji | Indecision: open and close nearly equal | In thin hours, where they are everywhere |
| Shooting star | Rejection of higher prices | With no resistance above to reject |
| Marubozu | Sustained pressure with no pullback | After an already extended move |
Look at the third column: in every case, what kills the pattern is context. Which is why “hammer equals buy” is not a rule, it is a headline.
The step that turns a pattern into a rule
A definition works when two people applying it to the same chart mark exactly the same candles. Compare:
| Vague | Measurable |
|---|---|
| “A hammer with a long wick” | “Lower wick at least twice the body, upper wick smaller than the body” |
| “At an important support” | “Within 0.5 ATR of the lowest low of the last 20 candles” |
| “With good volume” | “Volume above the 20-period average” |
| “In an uptrend” | “Price above the 50 moving average on the higher timeframe” |
How to test it in 30 minutes
- Write the measurable definition, with all four conditions from the table above.
- Scroll the history and mark the first 100 cases that meet it. Do not choose: take them all.
- For each, record what would have happened with a 1.5 ATR stop and a 2R target.
- Calculate the average expectancy in R.
- Compare against a baseline: do the same with 100 entries taken at random in the same instrument and session.
Step 5 is the one almost nobody does, and the only one that matters. A pattern with a 55% win rate sounds good until you discover that entering at random in that same session produced 54%. What you are looking for is not a high number — it is a difference against having no pattern at all.
Mistakes that inflate your results
- Hunting for the pretty examples on the chart instead of taking them in order.
- Changing the definition mid-measurement to include a case that “obviously counts”.
- Measuring on one instrument and assuming it holds for all of them.
- Forgetting costs: an edge of 0.05R disappears into the spread.
- Testing in a clean trending period and expecting the same in a range.
What to do with the answer
If the pattern clearly beats the baseline, you now have a rule for your plan. If it does not, the useful part is knowing now rather than after 200 trades. Either way, what turns a rule into results is executing it the same way every time: the Cortex Signal Kit delivers signals already defined with entry, stop and target, and Cortex Automation executes them without interpretation. For the full testing method there is the free ebook The Trading Journal Blueprint.
Do candlestick patterns work?
Some do, in specific contexts and specific instruments. As a general, isolated rule, no.
How many cases do I need to decide?
100 at minimum. With fewer, the difference you see is probably chance.
Which timeframe gives more reliable patterns?
Higher timeframes usually give fewer and cleaner signals, because there is less noise per candle.
Does volume matter on forex or index CFDs?
It is your broker's tick volume, not real market volume, so treat it as a relative reference and never as an absolute figure.
Can I automate a candlestick pattern?
Yes, if your definition is measurable. If you cannot write it as numeric conditions, you cannot test it either.
Measurable rules, not headlines. See the Cortex Signal Kit and Cortex Automation.
Disclaimer: educational content. The parameters are examples and should be validated with your own data. Cortex Next is not a financial service and guarantees no results. Trading involves risk of loss.
