The technical analysis literature on opening gaps in the market is filled with generalizations that may or may not be valid. Rarely do we see a hard statistical look at opening gaps. (A notable exception is the fine work of Jason Goepfert at the Sentimentrader site). So I thought I'd use Monday's example of a filled opening gap to explore the whole topic.
First off, let's establish what an opening gap means. I'm using SPY daily data and examining the disparity between where the index closed the previous day and where it opens today. We can attribute these gaps to news events overnight and the influence of overseas markets. The gap is also influenced by economic data that are reported prior to the New York open. Regardless of the cause, a large gap is an indication that market participants have shifted their valuation of stocks. When a gap is filled, it means that the initial revaluation was probably an overreaction. An unfilled gap suggests that the overnight events may have been of fundamental importance, driving a shift in investor perceptions regarding equities.
Going back to March, 1997 (N = 2375 trading days), the median opening gap--up or down--has been .29%. We have had 98 upside gaps of +1% or more; 83 downside gaps of -1% or more. The gap has been relatively small (less than .20% and greater than -.20%) on 854 occasions, or more than a third of the time. Of all days, 1760 (about three quarters of occasions) eventually touch the prior day's closing price during the day session, which means that any opening gap was filled.
The average size of an opening gap is highly sensitive to the market's overall volatility. When the VIX has been 20 or greater (N = 1382), the median size of the opening gap has been .39%. When the VIX has been under 20 (N = 993), the median gap has been .20%. Interestingly, the same proportion of gaps is filled under volatile and non-volatile conditions. That very strongly suggests to me that the odds of filling a gap of, say, .40% very much depend upon overall market volatility. The odds would be pretty good in a volatile market; much more remote in non-volatile markets. It might make sense, when analyzing gaps, to evaluate them as a proportion of recent daily range, rather than as an absolute price change.
We've had 452 days in which the VIX has been below 15. The median opening gap on those occasions has been .17%, with 341 filled gaps--that same 3/4 proportion. During those days, we've only had 28 occasions of an upside gap greater than .40%; 13 of these were filled. There were only 12 occasions of downside gaps less than -.40%; 8 of those were filled in the day session. It thus hasn't paid in low volatility markets to sell the open when we've had large gaps down, but it also hasn't been common that we've had large downside gaps in low volatility markets.
And when we've filled large upside gaps in low volatility markets (N = 13)? The following day in SPY has been up 10 times and down only 3. A filled upside gap has not necessarily been a bearish short-term indication for the market, contrary to common impression. More on this interesting topic to come, both here and on the Trading Psychology Weblog.
Added note (8:00 AM CT): Of the 28 occasions we've had large opening gaps (.40% or greater)in a low volatility market, the market was up from open to close 16 times, down 12 times, for an average SPY change of .08%--better than average. If, however, we look to the close of the following day, the market was up 18 times, down 10 for an above average gain of .20%. The average gain for the sample overall is only .01%.