"Mind the gap" may be a useful warning on airports and train stations, but is it any good in short term trading? As traders we have all heard this general trading wisdom that says 'gaps get closed'. But are they? And also, let's be more specific about when they get closed. Is it the next day, next week or next year a gap gets closed?
Let's first start with what we define as gaps on a daily chart:
- Upward gap on a daily chart: the intra day low of a trading day is higher than the intra day high of the previous trading day.
- Downward gap on a daily chart: the intra day high of a trading day is lower than the intra day low of the previous trading day.
Of course gaps may exist on every time frame, but in this case we are only looking at the daily time frame because that's where this gap-closing-trading-wisdom has its origin. And since trading is inherently short term, we are going to look at what happened within the next 10 trading sessions after a gap was formed. It is of no use to take this to several weeks or even months when trading things like futures or options. A short term trader is looking for an edge in the next few days. So is there a short term edge in trading gaps?
We are going to look at the S&P 500 ETF: the SPY. At this moment of writing, there are over 6300 SPY-trading days in my database. The first table shows that there have been 387 upward gaps during that period. Next in the table, you will find the numbers showing how many gaps have been closed and on which day.
In the terminology below, we define 'the 1st day' as the day that caused the gap. So 'the 2nd day', is the first day after the day the gap was created.
|upward gaps||closed the 2nd day||closed 3nd day||... 4rd day||... 5th day||... 6th day||7th||8th||9th||10th|
So 10 trading days following the upward gap, 237 (=78+48+39+20+18+12+12+4+6) of the 387 have been closed.
Let's now look at the downward gap:
|downward gaps||closed the 2nd day||closed 3nd day||... 4rd day||... 5th day||... 6th day||7th||8th||9th||10th|
In 10 trading days following the downward gap, 140 (=41+28+25+14+7+12+5+3+5) gaps of the 210 have been closed.
Also note that the total number of gaps is 597 (=387+210) which is 9,5% of all trading days in this (huge) sample. For the other 9 out of 10 trading days, you have to look for other setups, other than these gap statistics.
So what should a trader do the day after a gap has been formed? Expect gap closing or the opposite? The numbers above speak for themselves. Some statements can be done based on these numbers:
- 20% of upward gaps are closed the 2nd day.
- 20% of downward gaps are closed the 2nd day.
- 50% of upward gaps are closed within 5 days
- After 10 days 61% of upward gaps are closed.
- 50% of downward gaps are closed in 4 days.
- After 10 days 67% of downward gaps are closed.
This leads us to the following trading recommendation:
- The most favorable trade the next day after the gap would be to trade with the direction of the gap, not against it(!). This means: long-positioned after an upward gap and short-positioned after a downward gap. Note that it doesn't mean that price has to continue to move in the direction of the gap. It could also go sideways for multiple days, and then close the gap.
- The most favorable trade with a 10-day horizon after a gap is to trade against the direction of the gap. Gap closing is the most likely scenario within a 10-day period. This means: short-positioned after an upward gap, long-positioned after a downward gap. Obviously these positions should be taken a few days after the gap to eliminate some risk of being stopped out in the mean time
So is the gap-closing-trading-wisdom true or false? Since it didn't specify a time horizon, generally we could say the claim holds true, but maybe in a different way than most market commentators think of. Most main stream market commentators expect immediate gap closing the next session, but the statistics show us that is (very) unlikely to happen. The numbers clearly show that only 1 out 5 gaps is closed the next session and that 4 out 5 are still in place.
These numbers can not be easily compared to the numbers mentioned in The Encyclopedia of Chart Patterns by Thomas Bulkowski (it's in my bookstore). In this massive book on page 362 the author discusses several types of gaps, but in my opinion does not clearly specify how to distinct between these several gap types (another example of why technical analysis is not science). In his study, some types of gaps could take up to 168 days(!) on average to close (break away gaps).
However I do like his idea about looking at gap-behavior in bull and bear market situations. Unfortunately he did not specify the exact bull or bear market criteria. Still, The Encyclopedia of Chart Patterns is of course an excellent book and a must have for any trader.
The next thing I would like to see is how a European index behaves with respect to gaps. Because of the time-difference between these markets (US vs EU), gaps will come up on completely different days and may behave in a different way. Let's find out by having a look at the Dutch AEX. In my database, I have data going back to 1992 which comes down to 5359 trading days at this moment of writing.
|upward gaps||closed the next day||closed 2nd day||... 3rd day||... 4th day||... 5th day||6th||7th||8th||9th||10th|
|downward gaps||closed the next day||closed 2nd day||... 3rd day||... 4th day||... 5th day||6th||7th||8th||9th||10th|
The first very obvious difference with the S&P 500 is the number of gaps on the AEX. Over 16% of its trading days have gaps(!) To sum it up:
- 26% of upward gaps are closed the next session.
- 31% of downward gaps are closed the next session.
- 50% of upward gaps are closed in 4 days
- After 10 days 63% of upward gaps are closed.
- 50% of downward gaps are closed in 3 days.
- After 10 days 76% of downward gaps are closed.
- Gap closing has a bullish bias.
Even though the AEX has many more gaps, these percentages show great similarities with the S&P data. The conclusion stays the same: trade with the gap the first day, but expect a close within 10 days.