Candlestick patterns need context

Scan the NSE for every bullish engulfing candle and you will get a long list in any market, on any day. That is the problem. A candlestick pattern is a sentence fragment — two or three bars describing a short fight between buyers and sellers — and a fragment means nothing until you know where it was spoken. The engulfing candle that matters happens somewhere: at a level, against a stretch, after a decline. The one that happens mid-range, mid-drift, on thin volume, is shape without story.

This is why every candlestick scan in the pattern library here carries context conditions alongside the pattern flag. The pattern proposes; the context disposes.

Filter one: location

A hammer is a rejection — price driven down, bought back, sellers refused. Rejection only informs when there was something to reject from. The hammer at oversold scan demands the stock be genuinely stretched — RSI under 40 — before the hammer counts; the bullish engulfing at support wants the engulfing bar printed within reach of the 50-day average, a level with an owner base behind it. In Sift the pattern is a field like any other, so the context is just more conditions:

pattern is bullish_engulfing within 2 bars
  and rsi(14) < 40

The within 2 bars matters too: patterns are events, and an event from earlier in the week can still be fresh — a nuance offset-based screeners handle with duplicated clauses, if at all.

Filter two: the preceding trend

Reversal patterns require something to reverse — and the three-bar patterns carry their trend requirement inside the definition itself. A morning star is a down bar, a small indecision bar, and a strong up bar, in that order; the pattern flag only fires when that sequence printed, which is why its scan can afford to be bare. Single-bar patterns cannot lean on sequence, so their scans add the check explicitly: the shooting star only counts when it prints into strength, with RSI already above 60.

The neutral case makes the point best. A doji is pure indecision, and indecision is only information where one side had been winning — so the scan requires the stock above its 50-day average, with volume running above normal, before a doji is worth a look.

Filter three: volume

Patterns describe a fight; volume says how many showed up to it. An engulfing bar on twice normal volume is a verdict with attendance. The same bar on a sleepy session is easier to dismiss — half the pattern scans here carry a rel_volume condition for exactly this reason, the same filter that separates real breakouts from drifts everywhere else on the site.

Then make the folklore testable

Candlestick lore is old, confident, and mostly unexamined on Indian data. Here it does not have to be: every pattern scan carries the hit-rate replay, so "does a morning star after weakness actually lead anywhere on the NSE?" is an empirical question with a ninety-millisecond answer. Read the table with the usual discipline — count, payoff, holding window — and let the patterns that fail replay go. The ones that survive context and replay are few, which is rather the point: a screener's job is not to find you more candles, it is to find you fewer.