Testing Higher-Time-Frame Features for Short-Horizon EURUSD Forecasts

Summary

This article tests whether changes in higher-time-frame prices improve forecasts of EURUSD’s future close over a short horizon. It compares ordinary OHLC inputs, higher-time-frame price changes, and a combined feature set. The strategy premise is that a higher-time-frame trend tends to persist on lower frames, so trades aligned with that trend receive greater weight; the article also notes that this bias can leave traders exposed while waiting for higher-frame confirmation of a reversal.

In the reported modeling exercise, OHLC predictors performed best, while higher-time-frame changes showed weak relationships with the target and were removed during backward feature selection. Linear regression set the validation benchmark, and a Gradient Boosting Regressor was tuned through randomized search and local optimization, but neither tuning nor the selected model surpassed that benchmark on validation data. The author suggests overfitting as one possible explanation and notes that higher-time-frame indicators were not explored. The discussion is an empirical illustration rather than proof that multi-time-frame trading is ineffective; its conclusions are limited to the chosen data, target, features, models, and evaluation setup.

Key ideas

• The tested multi-time-frame premise gives more weight to trades aligned with a higher-frame trend.
• The experiment predicts a future EURUSD close using OHLC data, higher-time-frame price changes, or both.
• Higher-time-frame price changes had weak observed relationships with the target and were discarded by backward feature selection.
• Linear regression provided the validation benchmark, which the tuned Gradient Boosting Regressor did not exceed.
• The results are limited to the selected forecast setup, and other higher-time-frame features were not tested.

Tags

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