Charts Without Illusions · Lesson 3 of 5
Patterns and Pareidolia
The brain that finds faces in clouds is the same brain reading your charts — and it does not label its inventions.
Humans are pattern-finding machines with no off switch — the tendency is called pareidolia, and on charts it works overtime. Given random data, people reliably "see" trends, levels and shapes: flip a coin a few hundred times, chart the running total, and the result shows beautiful trends, double tops and clean support — in pure noise, where prediction is impossible by construction. Your eye cannot tell that chart from a real one. Sit with that: the shapes appearing is not evidence the shapes mean anything.
The classical pattern canon — head and shoulders, flags, triangles, double bottoms — has been studied academically, and the honest summary is: results are mixed, effects small where found at all, and hostage to definitions. Which exposes the deeper problem: most patterns are not defined tightly enough to test. How exact must the two shoulders be? Where precisely is the neckline? Two believers disagree on whether the pattern is even present — and an untestable claim isn't folklore-that-might-be-true, it is unfalsifiable, the lowest rank a claim can hold. Hindsight completes the trap: textbook examples are found after the move, while the failed near-patterns nobody screenshotted vanish from memory — survivorship bias, drawn in candles.
Now the honest complication — one family of levels is partly real: the ones everybody watches. Round numbers, prior highs and lows, widely-published levels. Not because geometry has power, but because coordination does: thousands of traders park orders at the same obvious places, and the clustered orders themselves cause reactions there. A self-fulfilling mechanism — genuinely existing, routinely overwhelmed by news, and carrying a lesson pattern-believers skip: the power lives in crowd behaviour at obvious prices, not in trendline artistry. The more artistic (subjective) your level, the fewer people share it, and the less coordination it can summon.
The test that cuts through all of it: could a stranger, given your written definition, mark the same pattern on the same chart — before the outcome? If yes, you have a testable hypothesis; lesson-three-of-the-backtest-course rules apply. If no, you have pareidolia with terminology. Almost everything in the pattern canon fails this test as commonly taught.