Testing Alternative Bar Types for FX Strategy Efficacy

Summary

This study tests whether activity-based and information-driven bars improve strategy efficacy over time bars using EURUSD quote ticks from 2022–2023. Because spot FX quotes lack consolidated traded volume, the authors show that several nominal volume and dollar bar variants collapse into tick-based equivalents. They therefore compare four distinct representations: time, tick, tick-imbalance, and tick-runs bars. Each is calibrated to produce a similar number of observations, helping isolate bar representation from sample-size effects.

The experiment applies RSI and Bollinger Band mean-reversion rules and an ADX/DI trend rule, labels entries with triple barriers, then evaluates a uniqueness-controlled random forest using purged cross-validation. Out-of-sample AUC values range from 0.42 to 0.55, and a permutation test on the best-looking result gives a p-value of 0.10. The authors conclude that better statistical conditioning does not automatically improve these strategies. Results are specific to the rules, quote-only data, and study period; the near-balanced quote direction also weakens the rationale for information bars based on trade-flow persistence.

Key ideas

• On quote-only spot FX data, several volume and dollar bar types reduce to tick-bar variants.
• Matching bar counts across families helps distinguish representation effects from sample-size effects.
• The study compares time, tick, tick-imbalance, and tick-runs bars with mean-reversion and trend-following primaries.
• Purged cross-validation and a permutation test find no reliable efficacy advantage for any tested family.
• Better-conditioned return samples do not necessarily create a stronger strategy edge.

Tags

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