Rolling Linear Regression for Intraday Trading Signals

Summary

The document describes a simple trend-following system built with linear regression. It introduces ordinary least squares and explains rolling regression as fitting a model repeatedly using a recent window of candles. In the example, the model relates each candle’s open and close prices. For the current candle, its opening price is used to predict its close: a forecast above the open triggers a long position, while a forecast below the open triggers a short position.

The example uses hourly Nifty data and reports decent performance across a historical period. The article does not provide enough detail to independently assess that claim, and its stated simulation dates are inconsistent. It also says slippage is excluded. The method is presented as an intraday strategy, but no robust out-of-sample validation, transaction cost analysis, or detailed risk results are given. The article frames rolling regression as a walk-forward approach and connects regression on a series to autoregressive time-series methods.

Key ideas

• Rolling regression refits a model on a moving window of recent observations.
• The example predicts a candle’s close from its open using a regression fitted to earlier open and close data.
• The trading rule takes a long position when predicted close exceeds the open and a short position when it falls below the open.
• The reported historical performance is limited evidence because slippage is omitted and the simulation dates are inconsistent.

Tags

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