Swing Point Detection

Swing point detection is the foundation of all harmonic and structure pattern detection in the ohlc-service. It identifies significant pivot highs and lows from a series of OHLC candles.

Algorithm

Constant: PIVOT_LOOKBACK = 5

A candle qualifies as:

• Swing high — its high is ≥ the high of every candle within 5 candles before and 5 candles after it
• Swing low — its low is ≤ the low of every candle within 5 candles before and 5 candles after it

Minimum candle requirement

PIVOT_LOOKBACK * 2 + 5 = 15 candles

Fewer than 15 candles returns an empty array — there is not enough surrounding context to confirm any pivot.

Output

findSwingPoints(candles: OhlcCandle[]): SwingPoint[] (in patternDetector.ts)

The result is an alternating list of highs and lows — a swing high is always followed by a swing low, and vice versa. If two consecutive pivots of the same type are detected, the more extreme one is kept.

Each SwingPoint carries:

Field	Description
index	Candle index in the input array
price	The pivot price (high or low)
type	'high' or 'low'
timestamp	Candle open time

Role in Pattern Detection

Every harmonic pattern (Gartley, Bat, Butterfly, Crab, Shark, Cypher) and every structure pattern (ABCD, Three Drives, Double Top/Bottom, Head & Shoulders) operates on the swing point list rather than raw candles. The swing points define the X, A, B, C, D legs used in ratio calculations.

See Markets — Pattern Detection Overview for the full pattern family map.