How reliable are candlestick patterns on a one-minute chart?

[Amanhasnoname]

Patterns like engulfing candles, pin bars, and dojis are popular on higher timeframes. On one-minute charts, a candle consists of a few dozen to a few hundred ticks, and its shape is heavily influenced by noise and spread.

Statistical studies of such patterns on short intervals typically find weak predictive power, often below the level needed to cover costs. The patterns appear frequently and their outcomes are close to random.

The shape of a one-minute candle also depends on the broker's feed, since the high and low come from the bid prices and different feeds have different extremes.

A scalper using candlestick patterns should test them objectively: define the pattern precisely, count occurrences, and measure the forward return after costs across many instances and several pairs.

If the results are not distinguishable from zero, the pattern adds nothing. If there is a small effect, check whether it holds out of sample and across different feeds.

Context can improve patterns, such as location near a level or a time-of-day filter, but these should be tested too.

Which candlestick patterns survive objective testing on the one-minute chart?