Testing Candlestick Clusters for Sequential and Return Patterns

Summary

This article uses k-means clustering to group daily GBP/JPY candles according to their high, low, and close relative to the open. It examines whether particular candle clusters tend to follow one another and whether returns after each cluster differ. The analysis separates data into an in-sample period and a later test period, applying the learned clusters to the later observations.

The reported transition table suggests that large moves may be followed by similar or smaller moves, while quieter candles also tend to cluster together. The article proposes examining next-day returns by cluster and comparing cluster-based returns with buy-and-hold, including a post-crisis period. The excerpt does not provide detailed numerical return results or establish statistical significance, so it does not support a firm trading conclusion. The author notes that daily forex boundaries depend on the chosen close time, intraday patterns may require volatility-cycle adjustments, and transaction costs should be included. Other possible extensions include more clusters, hierarchical methods, additional features, and multi-candle patterns.

Key ideas

• K-means can classify candles using their high, low, and close relative to the open.
• A transition table can reveal whether some candle types tend to follow particular prior types.
• Cluster-conditioned returns can be compared with buy-and-hold and checked on later data.
• The choice of daily close, transaction costs, and the number of clusters can affect the conclusions.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.