Please rotate your device to landscape mode to view the charts.

Charting by Machines

Journal: Journal of Financial Economics

Date: 20240301

Author: Murray, Scott; Xia, Yusen; Xiao, Houping

Abstract:
We test the efficient market hypothesis by using machine learning to forecast stock returns from historical performance. These forecasts strongly predict the cross-section of future stock returns. The predictive power holds in most subperiods and is strong among the largest 500 stocks. The forecasting function has important nonlinearities and interactions, is remarkably stable through time, and captures effects distinct from momentum, reversal, and extant technical signals. These findings question the efficient market hypothesis and indicate that technical analysis and charting have merit. We also demonstrate that machine learning models that perform well in optimization continue to perform well out-of-sample.

Link: Google Scholar

Background and Context

Research Question

This study tests whether machine learning can predict stock returns using only historical price data, challenging the Efficient Market Hypothesis which suggests this should not be possible.

Methodology

The researchers used a convolutional neural network with long short-term memory to analyze 12 months of historical stock price data from 1927-2022, optimizing the model on 1927-1963 data before testing on 1963-2022.

Significance

The study provides new evidence on the validity of technical analysis and charting, which remains widely used by investment managers despite academic skepticism.

ML-Based Strategy Generates Significant Excess Returns

  • Shows average monthly excess returns for portfolios sorted by ML forecasts (MLER)
  • Portfolio returns increase monotonically from -0.14% (lowest MLER) to 0.93% (highest MLER)
  • The strategy of buying high MLER stocks and selling low MLER stocks generates 1.08% monthly excess returns

ML Strategy Performance Persists Across Time Periods

  • Shows performance of ML strategy across different time periods
  • Strategy generates positive returns in most subperiods
  • Recent period (2015-2022) shows continued strong performance at 1.20% monthly returns

Strategy Remains Effective Among Large Stocks

  • Shows strategy performance across different stock size groups
  • Returns remain strong even among largest stocks
  • Top 500 stocks still generate 0.72% monthly returns

ML Strategy Captures Effects Beyond Momentum and Reversal

  • Shows strategy returns after controlling for known effects
  • Returns remain significant after controlling for momentum and reversal
  • ML strategy captures patterns beyond traditional technical indicators

Optimization Process Successfully Identifies Effective Models

  • Shows relationship between optimization and test period performance
  • Strong correlation (0.69) between in-sample and out-of-sample performance
  • Validates the model selection process

Contribution and Implications

  • Provides strong evidence against the Efficient Market Hypothesis by showing that historical price patterns can predict future returns
  • Validates technical analysis as a legitimate investment approach, explaining its continued widespread use by practitioners
  • Demonstrates that machine learning can successfully identify complex patterns in financial markets that persist through time

Data Sources

  • Returns Chart: Based on Table 3 showing portfolio returns sorted by MLER
  • Subperiod Chart: Based on Table 5 showing performance across different time periods
  • Size Chart: Based on Table 6 showing returns across different size groups
  • Control Chart: Based on Table 12 showing returns after controlling for other effects
  • Optimization Chart: Based on Figure 7 showing relationship between optimization and test period performance