Tuesday, November 21, 2006

Do Opening Gaps Tend to Fill?

The difference between the market's opening price and its previous day's close forms a gap on a bar chart. Does this gap tend to fill in during the next day's price action? That is, when we open with a gap lower or higher, do prices subsequently move back to the prior day's close?

For this investigation, I measured gaps as a function of the previous day's high-low range. This measures the gap relative to the prior day's volatility. Thus, an opening gap of two points that follows a day with a range of six points is measured as a 33% gap. The same opening gap of two points that follows a day with a range of ten points is measured as a 20% gap.

Assessed in this manner, we find that the average opening gap is 27% of the previous day's range going back to May, 2003 (N = 897 trading days) in the S&P 500 Index (SPY). That provides us with a benchmark for defining relatively large and relatively small gaps.

When upside gaps exceed 40% of the prior day's range (N = 99), 46 of them fail to close during the day session. When downside gaps exceed 40% of the previous day's range (N = 81), 40 of them go unfilled. Bottom line: approximately half of all large opening gaps don't fill during that coming day's action.

Conversely, when the upside or downside gaps are less than 40% of the last day's range (N = 717), only 144 of them go unfilled. When the gaps are less than 20% of the last day's range (N = 431), only 51 of these fill in. Stated otherwise, 80-90% of relatively small gaps will fill in during the coming day's action.

Interestingly, when we have a large opening gap to the upside, the *following* day's price change averages .18% (68 up, 31 down). When we have a large opening gap to the downside, the next day's price change averages .09% (48 up, 33 down). But when the opening gap is small, the subsequent day's price change averages only .03% (381 up, 336 down). In sum, large gaps in either direction tend to be bullish for the next day's price change.

When we measure gaps in this fashion, it takes relative changes in volatility out of the equation. The average size of the opening gap, defined as a function of the prior day's range, has been the same in 2006 as previously, despite our dramatic drop in volatility since 2003. Standardizing how we measure gaps provides us with a better sense for when gaps are large--and are less likely to fill--and when they're small--and more likely to fill.

26 comments:

MidKnight said...

That is outstanding research Brett. May I ask how you are scanning this? Just with excel or are you programming this in something like neoticker?

Kind regards,
MK

Brett Steenbarger, Ph.D. said...

Hi MK,

Thanks for the note. I find this kind of research very practical for guiding my strategy during the morning session. If I see a relatively small downside gap at the open and then see selling in the Market Delta dry up during the opening minutes of trade, I'll play for that gap-filling retracement--especially if other research has me bullish for the day.

I performed the research entirely within Excel, which is how I conduct all the research on the blog. I appreciate the interest--

Brett

John Wheatcroft said...

Great information - as a gap player I appreciate this data.

You say "Interestingly, when we have a large opening gap to the upside, the *following* day's price change averages .18% (68 up, 31 down)."
- Question - Is this for a stock that gapped up and then went down or for one that gapped up and continued to go up? I.E. one that closed higher than the open vs one that closed lower than the open.

Anonymous said...

Brett,
Excellent research - I find this data very important when looking to fade the gaps! I will definitely keep this in mind and start to chart the results myself (above/below 40% of the previous day’s range).

Chris

Brett Steenbarger, Ph.D. said...

Thanks, John. My research covers all the opening gaps, whether they were retraced or not. A big open gap to the upside seems to be associated with bullish action the following day.

Brett

Brett Steenbarger, Ph.D. said...

Hi Chris,

I appreciate the comment. I do think that there might be some good ways of handicapping the odds of movements after gaps based on this research and also the volume and distribution of NYSE TICK early in the AM trade.

Brett

Mike_Trader said...

Dr. Brett,
I've found similar results in my testing. It's interesting how much you learn when you challenge widely held assumptions. Here is a suggestion for your methodology. Using the prior day only for the range is subject to wide swings in gap size. I like to use the median of the 10 period True Range to assess what is normal movement for a stock. I find this works best because it eliminates large spikes that are not representative of a 'normal' day. Thanks and keep up the good work.

Mike

Brett Steenbarger, Ph.D. said...

Hi Mike,

Thanks for the excellent suggestion. That would certainly smooth out daily shifts in volatility. It's a promising area of research; I appreciate the feedback--

Brett

Anonymous said...

Hi Brett,

Really excelente post. I would like to translate this post and other post to publish in my website (www.especulacion.org) that is in Spanish, can I do it?.

Thanks in advance.

Jesus Perez

Brett Steenbarger, Ph.D. said...

Hello Jesus,

That is a *great* idea. I would be honored to have my posts translated on your site. Similarly, if you translate any of your own favorite posts into English, I would be happy to link to those from my personal site. The markets are truly global, and blogs should be that way also!

Brett

Anonymous said...

Hi Brett,

I just have discovered your Blog, Thanks for your hard work!

Almost a silly question when you say next day you mean the same day that the opening gap occurs, not the next day of the opening gap.

Just in case,
Thanks in advance,

DanielV

Brett Steenbarger, Ph.D. said...

Thanks for your note Daniel, and sorry for any confusion. If today's market open shows a large gap, tomorrow's price change (today's close to tomorrow's close) tends to be bullish. That's what my results were suggesting.

Brett

Anonymous said...

Hi Brett,

thanks for you effort to help us increasing our trading abilities.If i would have got this at my starting time as a trader i could save much money and avoid many bad emotions.
I can say that my trading skills are undoubtly better since i read your postings.
I use also Excel to calculate the odds for a system but i couldnt find a possibility to use this for so many stocks like you.
Thats means uncountable hours of work.
How you do this?
i do it like this:
I put the calculations in an Excel Sheet and then i put the data for every stock in this sheet with copy/paste.
How calculate the odds for so many stocks when you have an approach.

Pierrus

D TradeIdeas said...

Great article Brett. This is just the thing to model in The Odds Maker and determine how profitable a strategy based on gaps that fill can be.

If I could only get out from all the follow up from the recent Expo! It's all the more amazing that you consistently produce such great articles with your busy schedule.

Brett Steenbarger, Ph.D. said...

Hi Pierrus,

Great question; thanks for the note. I only trade the stock indices, so I don't try to calculate odds for all stocks. That would be onerous. I think that's why the Trade Ideas folks added their Odds Maker program to their service--it automates that process for you.

Recently, I've begun branching out to testing odds for ETFs. That is a very promising area of research.

Brett

Brett Steenbarger, Ph.D. said...

Hi David,

Yes, Odds Maker is very good at testing these kinds of technical patterns. Basically it just boils down to looking at market history, seeing what has worked, and then making the leap of faith that it will work at least during the next time period!

Brett

Anonymous said...

Thanks for the answer.

This may be old, or have answered already, what soft do you use in your trading platform apart of Market Delta?

Brett Steenbarger, Ph.D. said...

Hi Daniel,

My daily trading software includes e-Signal, RealTick, Trade Ideas, and Market Delta.

Brett

Anonymous said...

Great site Brett! So I am looking for someone who might be licensing or selling a trading system based on gaps. Do you know of any you might recommend?

Thanks!

Brett Steenbarger, Ph.D. said...

Thanks for the note, Happy1. I don't know of any gap systems for sure, but one source you might check on is Mike Bryant at www.breakoutfutures.com. If he hasn't developed something like that, he might point you in the right direction. Excellent resource for systems.

Brett

Audacity17 said...

Brett, inspired by you I have recently realized the power of spreadsheets to conduct my own breadth/sentiment research based on criterion I choose. Unfortunately, I realize my mastery of Excel is poor. What books do you recommend for increasing my knowledge? How did you learn to use them?

Brett Steenbarger, Ph.D. said...

Hi Audacity,

Any intermediate-level Excel book should do the trick. Make sure, when you look in the index, that the book covers the database functions within Excel: sort, filter, etc. Those are the ones I find most helpful. I really just learned by reading a basic text and playing around with it. It's definitely one of those things that becomes very easy once you get a routine down.

It's the one topic I'd like to do a seminar on some day: how to conduct your own market research to find an edge. It really has proven helpful in my own trading. Thanks for the interest--

Brett

UrbaneGorilla said...

Ran a backtest on the same matter about 2 weeks ago. I found that the relationship between the open and the high, low and close of the prior day factors in. (In effect, was yesterday an up day and did it peter out at the end of the day, or was it strong into the close). I also noted that whether yesterday's close was an up day compared to the prior day also affect's the movement of from the open. Needless to say, it gets complicated, but I can see generating a set of rules dependent upon what the action was in the prior two sessions.

Thx for the article UrbaneGorilla

Lover Dad said...

Brett, to clarify, you compare the previous day H/L or the O/C? To the close/open overnight gap.

JimRI said...

You noted today, April 29, 2009, that you would like to update this. That would be very valuable. This market may not be behaving like it was in 2006.

I have been trying to collect some history the crude way of printing out daily charts and using a pencil.

To me this information should reduce the ambiguity that I always feel in the AM which often keeps me from doing anything and missing some good runs.

Brett Steenbarger, Ph.D. said...

Hi Lover Dad,

I'm comparing high and low day session prices to the prior overnight gap; thanks--

Brett