What are chart fractals?

In mathematics, a fractal is a shape that looks similar at different scales - zoom in and you see echoes of the whole. In charting, the word is borrowed more loosely: a chart fractal is simply a recurring price pattern, a shape that has appeared many times across different assets and eras because the behaviour that produces it repeats.

Why patterns repeat

Prices are made by people (and, increasingly, by models trained on what people did). Fear, greed, hope and capitulation don’t change much from decade to decade. A market that has run too far pulls back the same way it did a century ago; a base that’s being quietly accumulated coils the same way whether it’s a 1950s blue chip or a 2020s token. The instruments change; the psychology rhymes. That’s why a normalised pattern from 1998 can be a genuine analog for one forming today.

What counts as a fractal in Chart Echo

Chart Echo doesn’t hunt for named textbook patterns (“head and shoulders,” “cup and handle”). Instead it treats whatever window you frame as the fractal and finds the historical windows whose shape is most similar - after normalising away absolute price and size. This is more flexible than a fixed pattern library: you’re not limited to the handful of shapes someone decided to name, and the match is grounded in geometry rather than labels.

Scale-invariance: the useful part

Because the matching ignores absolute price, a $5 stock and a $5,000 index can be “the same pattern.” A move that took three weeks and one that took three months can rhyme if their shapes line up (with optional time-warping). That scale-invariance is what lets a relatively small amount of distinct human behaviour generate a huge, searchable library of precedents.

The honest caveat

Pattern similarity is real evidence, but it is not destiny. Two charts can look identical and resolve differently because the world around them differs - liquidity, rates, news, regime. The value of fractals isn’t that they predict; it’s that they let you replace a vague feeling (“this looks familiar”) with a measured distribution of what familiar situations actually did next. Read more on how to interpret that distribution in reading historical analogs.