Pattern probability, honestly

This is the lesson that closes the patterns track, and it argues partly against the rest of it.

Chart patterns are useful. They are far less reliable than they are usually presented as being, and the gap between those two statements is where most pattern-based trading fails.

Why published hit rates are inflated#

Every number you have seen quoted about a pattern's success rate is affected by at least one of these.

Selection after the fact. Patterns are identified retrospectively, on charts where the outcome is already visible. On a historical chart, a head and shoulders is obvious. In real time, you are looking at a possible formation that has not finished.

Definitional flexibility. How symmetrical must the shoulders be? How deep is a valid pullback? Loose definitions let a researcher include the ones that worked and exclude the ones that did not, usually without meaning to.

Survivorship in the teaching. Every example in every book is a chart where the pattern worked. Nobody publishes the hundred times the same shape appeared and nothing followed.

Costs excluded. Many published figures ignore spread, brokerage and slippage entirely. A pattern with a small theoretical edge can be net negative after real costs.

Regime dependence. A study covering a trending period will find continuation patterns work well. The same study on a ranging period finds the opposite. Neither result generalises.

What patterns genuinely provide#

Strip away the inflated claims and something real remains.

A compact description of what happened. "Price tried twice and failed" is genuine information, efficiently encoded.

A defined invalidation. This is the underrated one. A pattern tells you exactly where you are wrong — the neckline, the pattern extreme, the range boundary. A trade with a clear invalidation is manageable regardless of hit rate.

A structure for risk. Knowing your stop before entry lets you size the position properly, which matters more than the hit rate itself.

A shared vocabulary. Because everyone sees the same shapes, positioning concentrates around them — which is itself a tradeable fact, as the failed-breakout lesson showed.

Notice that none of these require the pattern to be predictive. They are all about structuring a trade, which is what patterns are actually good for.

Expectancy, not hit rate#

The question "does this pattern work?" is the wrong question. The right one:

Expectancy = (win rate × average win) − (loss rate × average loss)

A pattern with a 40% hit rate and a 3:1 reward-to-risk is strongly profitable. A pattern with a 70% hit rate and a 1:3 payoff is a slow disaster.

This reframes everything. You are not looking for patterns that are right often. You are looking for patterns where being right pays more than being wrong costs — and where you can define both in advance.

It also explains why the hit-rate obsession is misplaced. Hit rate is one of four inputs, and the one most affected by all the biases above.

Testing one for yourself#

You do not need backtesting software. You need records and patience.

Define it precisely, in writing, before you look. "Bullish engulfing" is not a definition. "Second candle's body fully covers the first's, second candle closes in the top third of its range, after at least three consecutive down closes" is.

Log every occurrence, including the ones you skip. This is the step everyone omits and the one that removes the survivorship bias. If you only log the ones you traded, you have rebuilt the same distorted sample the books gave you.

Record the outcome against a fixed rule. Same stop and target logic every time, so you are measuring the pattern and not your management.

Fifty occurrences before you conclude anything. Twenty is noise, as the journal lesson said.

Include costs. Every time.

Where this leaves the track#

Patterns are a way of reading and structuring, not predicting.

Use them to describe what happened, to define where you are wrong, and to size accordingly. Do not use them as a reason to expect an outcome at a rate you read somewhere.

The traders who do well with patterns are not the ones who know the most of them. They are the ones who have tested a few honestly, know their real expectancy including costs, and size accordingly — which is a much duller skill than pattern recognition and a far more durable one.

Check yourself

0 of 4 answered

1. 1.Why are published chart pattern hit rates systematically inflated?

   Researchers deliberately falsify themRetrospective identification, flexible definitions, survivorship in teaching material, and excluded costs all push the number upThey are measured on foreign markets onlyThey were accurate historically but markets changed

2. 2.Which is the more profitable setup?

   70% hit rate with 1:3 reward-to-risk40% hit rate with 3:1 reward-to-riskThey are equivalentCannot be determined without more information

3. 3.When testing a pattern yourself, why must you log occurrences you did not trade?

   For tax record-keepingLogging only what you traded rebuilds the same survivorship bias the textbooks haveBrokers require a complete logIt is not necessary if you have enough trades

4. 4.What do chart patterns genuinely provide, even if they are weakly predictive?

   A reliable forecast of the next moveA defined invalidation, a structure for sizing risk, and a compact description of what happenedConfirmation that a trend will continueNothing of value

What to take away

• Published hit rates are inflated by retrospective selection, loose definitions, survivorship and excluded costs.
• Patterns provide a defined invalidation and a structure for sizing — not a forecast.
• Expectancy beats hit rate. 40% at 3:1 beats 70% at 1:3.
• Test yourself: precise written definition, log every occurrence including skipped ones, fixed rules, 50 samples, costs included.
• Expect to find one or two setups carry everything.
• The durable skill is knowing your real expectancy, not knowing more patterns.