In trading, understanding range—the magnitude of price movement within a specific period—is essential for gauging market volatility and planning effective strategies. Unlike traditional volatility metrics such as the VIX or Implied Volatility (IV), which reflect market sentiment and its influence on option prices, range measures the absolute movement of the underlying asset itself, offering a direct view of price dynamics. Traders often rely on tools like the Average True Range (ATR) to analyze historical ranges. In this experiment, we take a step further by predicting the next day's range using advanced machine learning techniques.

What is Range Prediction?

Range prediction involves forecasting the span of price movement for the upcoming trading session. This insight can help traders anticipate market behavior and make more informed decisions. Unlike ATR, which summarizes past data, this experiment focuses on predicting future range using historical price and volume data.

Model Overview

We employed a Multilayer Perceptron (MLP) classifier, leveraging a classification using regression approach. This method combines the precision of regression with the categorical simplicity of classification to provide actionable predictions. The model categorizes the predicted range into three classes:

• HIGH
• MEDIUM
• LOW

Performance

The range prediction model demonstrated an accuracy of 68%, indicating a solid baseline for practical use. While not perfect, this level of accuracy provides meaningful insights, especially when combined with other trading signals.