Stats › Intraday and market microstructure › Opening gaps
Opening gaps and how often they fill
Gap fill is one of the most confidently repeated numbers in retail trading. Depending on who you ask, gaps fill 70% of the time, or 80%, or 92%, and it's always stated as though somebody counted. I went looking for the person who counted.
What I found is that no exchange, no clearing house, no regulator and no peer reviewed paper publishes a gap fill rate for the S&P 500 or its futures. The figures in circulation come from private studies and broker education pages, most of which don't state a definition at all. Two of them do state one, properly, with instruments and date ranges, and they're on this page with that label attached.
So this page does two things. It reports the part that is genuinely measurable, which is the frequency and size distribution of opening gaps and the overnight versus intraday split of returns, and it shows you why a single headline gap fill number is close to meaningless. The same dataset gives 43.3% or 60.3% depending only on when you stop the clock, and 77.8% or 8.2% depending only on how big the gap was.
In a 2,791 day sample of E-mini Nasdaq 100 futures from 2015 to 2025, every single day opened away from the prior regular session close, because the contract trades nearly around the clock. 60.3% of those gaps were fully filled by the close, but only 43.3% by 10:30, and the rate ranges from 77.8% for gaps under 0.3 times ATR down to 8.2% for gaps over 1.2 times ATR. That's one private study. A second, on SPY over six month windows, gets roughly 60% overall and the same steep decline with gap size. No exchange, regulator or peer reviewed source publishes a gap fill rate at all. What is properly published is the overnight split: the New York Fed measured +3.6% of the S&P 500's +5.9% annualised log return from 1998 to 2020 accruing between the 16:15 futures close and the 09:30 open.
Members only
The rest of this page is for members
Below this point there are 8 sections, 1 chart, 1 table and 5 named sources, roughly 2100 words of it. Every figure carries the source it came from and the date the data is from.
You can keep browsing every statistic in the library for free. The intro and the summary are always open.