Chart patterns: what they actually tell you (and why we don't let them vote)

Head and shoulders. Double top. Ascending triangle. If you've spent any time around trading content, you've seen someone draw one of these on a chart with total confidence, like it settles the argument. Sometimes it's genuinely useful. Sometimes it's someone finding a shape they were already looking for.

We built a detector for these patterns because they're a real, recognizable part of how price behaves. But we also tested whether they should influence our signals — and the honest answer changed what we shipped.

What a chart pattern actually is

Strip away the names and every classic pattern is describing the same underlying thing: a story about who's in control and where that control might be running out.

A double top is two failed attempts at the same price level — buyers pushed twice, got rejected twice, and the second failure tends to convince the crowd that still holds from the highs to give up and sell. A head and shoulders is a rally, a bigger rally, then a weaker rally — three attempts with the middle one strongest, which often means the buying pressure peaked and has been fading since. A triangle is compression — price coiling into a smaller and smaller range as buyers and sellers reach a temporary standoff, with the eventual breakout direction taken as a signal of who blinked first.

None of this is magic. It's a shorthand for "here's how the tug-of-war has been going," read off the shape it left behind.

Where people fool themselves

The pattern isn't real until the price action confirms it. A head and shoulders isn't a head and shoulders until the neckline actually breaks — before that, it's just three bumps that might turn into anything. A triangle that never breaks out isn't a signal, it's just consolidation that happened to look tidy in hindsight.

This is where a lot of chart-pattern content quietly cheats: it shows you the pattern after the move already happened, with the "obvious" breakout circled in red. Looking backward, everything looks obvious. The actual test is whether you'd have traded it with the same confidence before the confirmation candle closed — and most people, honestly, wouldn't have.

The other common mistake is pattern-shopping: staring at a chart until you find a shape that fits the trade you already wanted to take. With enough imagination, almost any chart contains a pattern. That's not analysis, that's confirmation bias with extra steps.

What we actually did with ours

Pairvue's chart-pattern detector looks for the same swing structure our support/resistance logic already tracks, and only calls a pattern real once its confirmation condition is met — a broken neckline for a head and shoulders, not just the shape leading up to it. You'll see it in the Analytics view on a pair when one shows up.

Here's the part that matters more than the detector itself: we tested whether counting a detected pattern as a vote toward the overall signal actually improved anything. Across a wide set of pairs, it didn't move the needle in a way we'd trust — sometimes a little better, often about the same or worse, with no consistent edge we could stand behind. So it doesn't vote. It's shown to you as information, not folded into the score.

That's the same rule we apply everywhere else: if we can't show the evidence, we don't ship the claim. A feature that looks compelling on a chart isn't the same as a feature that's earned a vote in the number that decides whether you get a signal.

How to actually use them

Treat a chart pattern as context, not a trigger. It's most useful as a way to describe what's already happening at a level you'd be watching anyway — not as a reason to enter a trade on its own. If a pattern shows up at a support or resistance zone your other reasoning already flagged, that's worth noting. If it's the only reason for the trade, be more skeptical of yourself than usual.

And wait for the confirmation. The unconfirmed pattern is a hypothesis. The confirmed one is still just a shorthand for structure you could describe in plain language if you had to — which is a good test, honestly. If you can't explain why a pattern matters without naming it, you probably don't need the name.