Cloud Trading System That Switches Between Trend and Mean Reversion

Summary

This project describes a cloud based automated system for WTI futures that uses machine learning to classify market conditions as trending or ranging. Several models vote within separate trend and range groups; when the groups disagree, their confidence scores determine the regime. The selected regime then activates a simple trend following or mean reversion approach, using Z score thresholds for trend entries. A five minute process refreshes the assessment, while broker data, queued signals, order acknowledgements, and timeout handling connect the strategy to execution.

The authors report 74% out of sample performance, but the evidence comes from only six months of one minute bars, and the full model details are withheld. The article also describes practical issues in cloud resource management, integration, and testing, including the cost of relying on a live broker connection without a market simulator. The reported performance is not enough to establish robustness or live profitability.

Key ideas

• The system uses machine learning votes to classify conditions as trending or ranging.
• It applies trend following or mean reversion logic according to the selected regime.
• Model confidence scores resolve cases where trend and range classifications conflict.
• Broker messages and order timeouts form part of the execution architecture.
• The reported out of sample result is based on a short sample, and model details are limited.

Tags

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