Quantile Thresholds for Forecast-Based Spread Trading

Summary

This strategy forecasts the future value of a spread between cointegrated assets, then compares the forecast with the current spread to generate trades. The document describes three approaches: trading predicted spread returns directly, following spread momentum, and trading forecast deviations. It reports that the deviation approach performed best in the authors’ empirical comparison, aiming to capture abrupt spread moves.

The forecast uses an automatically selected ARIMA model, with the Akaike Information Criterion used to choose model parameters. Entry thresholds come from separate quantiles of positive and negative spread changes measured during a formation period; the authors recommend considering the 10th or 20th percentiles. The document illustrates the method and gives a crude oil and gasoline futures example, but provides no detailed performance statistics. Results depend on cointegration and forecast quality, and optimizing thresholds directly for profit may introduce data snooping.

Key ideas

• The strategy forecasts a spread and trades when the forecast differs sufficiently from its current value.
• The documented empirical comparison favored forecast deviations over direct return prediction or spread momentum.
• An automatically selected ARIMA model forecasts the spread, with AIC guiding model selection.
• Separate quantiles of positive and negative formation-period spread changes set entry thresholds.
• The method assumes a suitable stationary spread and its example does not establish general performance.

Tags

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