Objective research to aid investing decisions

Value Investing Strategy (Strategy Overview)

Allocations for May 2024 (Final)

Momentum Investing Strategy (Strategy Overview)

Allocations for May 2024 (Final)
1st ETF 2nd ETF 3rd ETF

Intraday Stock Price Momentum and Reversal Trading

| | Posted in: Momentum Investing

Are there profitable intraday stock price momentum and/or reversal strategies? In his January 2017 paper entitled “Intra-Day Momentum”, Oleg Komarov examines the profitability of intraday times series (intrinsic or absolute) and cross-sectional stock price momentum and reversal strategies. Time series strategies involve predicting the behavior of a stock based on its own past return. Cross-sectional strategies involve predicting the behaviors of equally weighted groups of stocks sorted into tenths (deciles) based on their respective past returns. Specifically, he segments the trading day into half-hours and overnight and then examines whether returns for past half-hours predict returns for future half-hours, focusing empirically on whether: (1) returns during 9:30-12:00 predict returns during 15:30-1600; and, (2) returns during 9:30-13:00 predict returns during 13:30-15:30. He conducts purely statistical tests and tests based on trading strategies. He concludes with consideration of overnight returns. Using cleaned data for a broad sample of U.S. common stocks (excluding microcaps) during January 1993 through May 2010, he finds that:

  • Over the sample period, annualized average stock returns are:
    • -9.2% during 9:30-10:00.
    • -5.9% during 10:00-10:30.
    • In the range -1.4% to 2.5% for half hours during 10:30-15:30.
    • 8.9% during 15:30-16:00.
    • 20.1% during 16:00-9:30.
  • Statistical tests for time series relationships between the intraday intervals specified above show that:
    • The 15:30-16:00 return relates positively to the return during the preceding 9:30-12:00.
    • The 13:30-15:30 return relates negatively to the return during the preceding 9:30-13:00.
  • However, statistical time series predictability does not translate to profitable time series momentum trading strategies, because morning losers tend to appreciate faster than morning winners during the afternoon, especially during the last half-hour (suggesting a cross-sectional effect).
  • Equally weighted portfolios formed from decile stock sorts based on 9:30-12:00 or 9:30-13:00 returns exhibit U-shaped afternoon return patterns. Both extreme morning winners and extreme morning losers tend to earn positive returns during the afternoon, especially during the last half-hour.
    • For example, sorting on 9:30-12:00 returns, the winner (loser) decile portfolio generates average gross daily return 0.077% (0.062%), translating to an annualized gross 19.4% (15.6%), during the last half-hour. Decile portfolios 2-9 mostly generate less than half those returns.
    • The effect is robust to stock characteristics (size, trading volume, liquidity, tick size, volatility and skewness), day of the week and variations in ranking and holding periods.
    • However, there appears to be structural break with the introduction of decimalization in 2001. Morning winners outperform morning losers before 2001, but the reverse holds thereafter (with the winner portfolio unimpressive).
  • Including the overnight return when sorting stocks into deciles suggests a profitable reversal strategy, going long (short) extreme losers (winners) during the afternoon. Such a hedge portfolio generates average gross daily return 0.06% to 0.07% during the afternoon.

In summary, evidence indicates that there are profitable intraday stock price momentum and reversal strategies as measured by gross average returns.

Cautions regarding findings include:

  • As stated, performance data are gross, not net. Accounting for daily entries and exits across large numbers of stocks would generate material trading frictions that may well produce net losses. Frequent traders may, however, consider intraday patterns when executing stock trades for other reasons.
  • Experimenting with different past return and future return measurement intervals in search of significant links introduces data snooping bias, thereby overstating expectations.
  • Intraday collection/processing/trade execution across a large stock universe is beyond the reach of many investors, who would bear fees for delegating to an investment/fund manager.
  • The sample ends nearly seven years ago. Fresher data may be instructive. Results for a modest sample at the stock market level (“Recent Intraday U.S. Stock Market Behavior”) differ overall and vary considerably by calendar month. 

For related research, browse:

Daily Email Updates
Filter Research
  • Research Categories (select one or more)