Module 3 · Lesson 12 of 13 · Reading price · About 5 minutes
Why Chart Patterns Fail: Base Rates, Hindsight and Context
On this page
Before you read: Breakouts, False Breakouts and Retests in Price Action
What you will learn
- Why a named pattern is usually identified only after it has finished.
- What a base rate is and why it matters more than memorable examples.
- How hindsight bias, selection and context make patterns look more reliable than they are.
Why do patterns look so clear in books?
When you see a perfect head and shoulders in an example, who chose that example?
Textbook examples are chosen because they look clean and finished. The many shapes that started the same way and then did something else are rarely shown. That is a form of survivorship bias: the examples that “worked” survive into the book.
What is a base rate?
Out of every time a shape like this appeared, how often did it do what the name suggests?
A base rate is how often something happens across all cases, not just the memorable ones. For a pattern, it would mean counting every instance that met a written definition, in advance, and recording what followed. Without that count, a pattern’s reputation rests on examples. Even with it, the rate depends on the market, the period and the exact definition, and it can change.
Pattern outcomes diagram
Marks: stipulated outcomes for this illustration, not validated pattern-success rules.
Hindsight picks the cases that worked; the rule finds them all.
Illustrative diagram constructed for teaching. Not real market data.
Selected remembered cases are 5/5 marked (100% within that selected set); the invented complete grid has 38/100 marked (38%) and 62/100 not marked. These different denominators illustrate selection bias, not a measured pattern success rate. No real detection/outcome sample is supplied.
Why is a pattern only clear afterward?
At what point could you first say a double top was a double top?
Most patterns are confirmed by a later move. Under the common definition used in chart patterns, a double top is only complete once price has turned down from the second high and broken the low between them. Before that, the same bars could become a range, a flag or a breakout. Hindsight bias makes the finished shape feel as if it was obvious all along.
What role does context play?
Would the same shape mean the same thing before a major data release and on a quiet afternoon?
Patterns form inside larger conditions: the trend on a higher timeframe, scheduled news, liquidity and volatility. A shape that appears during a thin holiday session is not the same event as one that forms on heavy volume. Patterns are a vocabulary for describing price. They are not rules that markets follow.
What this does not tell you about patterns
- The base rate for any pattern in any market. That requires a defined, tested count.
- Whether a pattern forming now will complete.
- Which context matters most. That changes over time.
- Nothing here suggests buying, selling or holding when a pattern appears.
Method in Practice: 40 invented cases, counted
Fictional example. John, Harbor Index futures, the prices, times, costs and results are all invented for teaching. This shows how a written plan can be built and tested. It is not a recommendation, and nothing here says how often this idea works.
Try it on a real chart. Test this idea yourself on past sessions with bar replay. Keep your own count, including the days with no trade.
The idea. A base rate counts every case a rule finds, not the memorable ones (this lesson). John writes a fixed "bull flag" rule, runs it over his test sample, and counts all 40 cases. Invented numbers: this shows how to count, not how the method performs.
Three outcomes
-
Remembered: the 5 cases he'd have shown a friend. All 5 reached the target
-
Counted: 40 cases. 26 completed: 11 reached the target, 13 hit the wrong-if, 2 closed at the time limit. 14 never completed
-
After costs: the 26 trades net +$15 in total, about +$0.58 per trade
Show the count
John's rule: a rise of at least 1.00 in 6 or fewer 30-minute bars, then 3 to 8 bars drifting down by no more than half the rise, then a close above the drift's high (completion). Wrong if a close below the drift's low. Target: the completion price + half the rise. Time limit 14:30.
| Group | Cases | Average result after costs | Total |
|---|---|---|---|
| Completed, reached the target | 11 | +$110 | +$1,210 |
| Completed, hit the wrong-if | 13 | −$95 | −$1,235 |
| Completed, closed at the time limit | 2 | +$20 | +$40 |
| Never completed | 14 | $0 | $0 |
| All 40 | +$15 |
What the five remembered cases would suggest: 5 of 5. What the count shows: 11 of 26 completed trades reached the target, and the total after costs was close to zero.
Test it yourself
- Write the pattern rule as numbers, and let the rule (not your eye) find the cases.
- Count the cases that never completed. They're part of the base rate.
- Work out the result after costs, per trade and per case found.
Where it breaks
- 40 cases is a small sample. A different 40 could look very different.
- The rule's exact numbers decide which cases count.
- The base rate in one market and period may not hold in another.
Calculators: expectancy
Reflection question: For the pattern you trust most, do you know how many cases you've counted, including the failures?
Check yourself
- Why might the examples in a pattern book overstate how often a pattern works? Hint: which examples get chosen?
- What would you need to measure a pattern’s base rate fairly? Hint: a definition written before you look at the results.
- Name two kinds of context that could change what a pattern describes. Hint: timeframe, news, liquidity.
- Does the grid’s 38% establish a real pattern success rate? Why can the five remembered cases show 100%?
Reflect on your own process
- Is the pattern definition in your plan written precisely enough to count every case, not just the memorable ones?
- When you review past charts, how do you record the cases that started like a pattern and then did something else?
- Which context notes, such as session, news or timeframe, does your journal capture each time you mark a pattern?
- For the pattern you trust most, do you know how many cases you've counted, including the failures?