Framing a pattern window

When you run a search, Chart Echo matches the pattern window you frame - typically anywhere from about 30 to a few hundred candles. That choice quietly decides what “similar” means, so it deserves a moment of thought rather than a reflexive default.

Short windows vs long windows

A short window (≈30–60 candles) captures the immediate setup - the flag, the thrust, the shakeout. It finds lots of matches and reacts to recent action, but it’s more sensitive to noise: tiny wiggles can dominate the shape. A long window (≈120–240 candles) captures broader structure - the whole base, the entire trend leg. It finds fewer, more meaningful matches but responds slowly and can blur a sharp near-term signal.

Practical starting points

• Tight (≈30): a single sharp setup - a breakout day, a reversal, a momentum thrust.
• Short (≈60): a flag or a multi-week consolidation resolving.
• Medium (≈90): the default for most swing setups; enough context without drowning the signal.
• Long (≈120): a full base or a complete trend leg.
• Wide (≈240): macro structure and cycle position.

The overfitting trap

It’s tempting to shrink the window until the matches look uncannily perfect. Resist it. Below a certain length you’re matching noise, and “perfect” matches of noise tell you nothing about the future. If a tiny window is the only way to get agreement, that itself is a signal: history doesn’t strongly rhyme with your setup right now. More on interpreting weak agreement in reading historical analogs.

Iterate honestly

It’s fine to try two or three window lengths to see whether the conclusion is stable - a robust signal tends to survive small changes in framing. What’s not fine is reframing until you get the answer you wanted. Frame to capture the setup, not to manufacture a verdict.