RM and Me got interested on a peculiar topic and started researching on this. The data extraction, analysis and everything belongs to Rm. "Yours faithfully" as always, is trying to grab the honors by presenting it. :)
- One of the most feared elements for a trader is gaps. The more experienced ones always try to close the position towards End Of Day as a measure of caution. What if there is a news based gap up, what if there is a global sell off based gap down - these worries really haunt the amateur trader. Carrying the position is always a Catch-22 for the prospective trader. We wanted to analyse the problem to the bare bones and shatter or reinforce the myths. :)
- We have taken NF data for the last 6 months - from May 2011 till date. It has a total of 155 trading days. Not big - but a reasonable sample. During the period, NF has fallen from 5800 to 4700 - a fall of approximately 1100 points - so it is mostly a bear market phase. Secondly, any gap less than 20 points - we considered it as flat, so as to reduce the noise.
- Firstly, please find the datasheet here.
- One of the most striking aspects that occur immediately is - out of 155 trading days - there were 38 gap ups, 40 gap downs and 77 flat open days. That is approximately 25% each of gap up and gap downs and 50% flat open days. Well, there are two ways to look at this. Firstly, I can see people saying it is useless to carry a position - because 50% times it does not pay to hold it. Secondly, I can see people who see no harm in carrying a position because it can hurt them only 25% times. ( 50% flat and 25% gap in their direction - do not hurt them. :) ). Take your own call, and trade with conviction.
- We tried to zoom it down to one more level - and analysed the gaps in terms of weekdays. Some interesting things came out.
- Firstly, out of 38 gap ups, 20 occurred on a Friday or Monday. So, clearly Weekends are marked for news driven gap ups. Or should we say news is made to order on a weekend? The probability of a Monday gap up increases a lot more when there is a Friday gap up. Further, Monday has the least flat days. So, contrary to the general belief, that it is better to go home empty on a weekend, it pays to hold a position over a weekend - if you get the direction right. The statistic also proves Girish's Statement that "Rallies occur in the night and weekends and Falls Occur in Broad Day light." :)
- It does not pay at all to carry a long position on a Wednesday evening. On a Thursday, we have only 3 gap-ups so far and 25 flat to negative days. So, buying a long on a Wednesday evening is strict no - according to this stat. :)
- Similarly, Shorting on a Thursday evening does not give any returns. We have only 5 gap downs and 27 flat to positive days. So, staying a short on a Thursday evening does not help much.
- 36 of the 77 flat days occurred on Wed and Thursdays. Clearly, carrying positions in the middle of the week does not pay much.
- We have added the day trading range as well - and analysed the ranges in terms of weekdays. Some more interesting things came out.
- Though Friday opens gap up on most days, it closed flat to negative on 23 out of 32 - approximately 75% of times. So, the concept of profit booking on a Friday is a real thing. :)
- Thursday's have the most positive trading range. Though, they open flat to negative on most days, they close down only on 9 out of 28 - 33% of times. So, Thursday buy at open should be generally good.
- I am sure you are disappointed. I know you might have come here expecting some analysis on planetary influences and such fancy stuff. I humbly accept that the title is a misnomer. But, in these days of race for attention grabbing - in these days of information over flow - I needed to resort to this to grab your attention.:)
- At VTrender, we always look for simplification of trading system. As with Maths and Chess, in trading also the simplest solution is always the best, the most elegant and effective. Hope this analysis helps you in simplifying your trading methods. Please offer your feedback. Ofcourse, feel free to make your own deductions from the data sheet.
Posts
13 comments:
I am the first
Kris, RM,
Good post. Makes good reading.
Hope we have a follow up post one month from now with the " I told you so..." :)
You never fail to stimulate our mind or tickle our imagination by your way off the beaten path perspective. Heartfelt gratitude.
@Shai
How did you guess it man? A weekly series of "I told you so" posts already in the pipeline. :)
Thanks RM.
For me Forward testing works better than back-testing :)
Dear Kris and RM,
Thanks for the wonderful post, will try to add more data, using your sheet.
Thanks
Wonderful post ...... kris and RM keep it up
Sharad
Hi
Good post. I felt that while forming conclusions about the behavior of each day, the sample size of 155/5 = 31 trading days each was a bit small. I tried to broaden the same research by looking at what numbers would result if data was taken for the last 10 years, last 5 years, and the last year. While the most recent results are the most relevant, they also suffer from a small sample size. So a comparison between recent and long term results should be helpful. To not distort things in the case of a low value of Nifty, instead of taking a 20 points as the 'flat' condition, I took it as 0.4% of the previous close. The datasheet is uploaded here -
http://www.4shared.com/file/fosIu_lM/QE13_-_Gaps_-_VT.html
My thoughts regarding some of your conclusions -
Some of the numbers that appear like an 'edge' over the sample size of 155 days, seem to disappear when taken a sample of the last 5 years. For instance, while in the last year only 11% of Gap Ups occurred on Thursdays (supporting your "Doesn't pay at all to carry a long position on Wednesday evening" point) the data over the previous 5 or 10 years has roughly similar values for all days. (All in the 15-24% range). Same is the case for the hypothesis - "Shorting a Thursday doesn't pay" - While in the last year only 16% Gap downs occurred on a Friday (and even fewer 13% on a Wednesday), these distortions go away when seen over a larger window. Occurrence of Flat days over a long time also seems more or less uniform over the five days.
Fridays, that have been down ranged this year (79% days were either flat or down), is similar to the rest of the days over a long term (62% days flat or down)
I feel that a slight tendency does exist where you point it out, but that tendency, over a long period seems too slight to, of itself, constitute an edge.
It is arguable that these numbers aren't to be interpreted in a vacuum, but have to be looked at in the context of the kind of market that is prevailing. So, while a bear market lasts, the most recent measures of seasonality are more relevant than those seen from the previous 5 years. But then I think there are a few layers of complexity that will get added to the way we draw our conclusions.
Just my thoughts after viewing the analysis. I'm a beginner and my intention is obviously not to criticize.
Would love to hear your thoughts on this.
As an aside, on 'attention grabbing', after looking at the title of the post as 'Astro' perspective, I almost didn't read the post. Was pleasantly surprised to not find anything related to 'planetary influences' :)
Cheers.
@tradesmyth,
thanks for a valuable feedback.
Yes, whatever you stated is valid. And on a sufficiently longer time frame - everything evens out. As it is - trading itself is a zero sum game - isn't it? :)
@ Tradesmyth,Kris,
Thank you for your feedback and the spreadsheet.
From your sheet, the biggest percentage across days seem to be 50-60 % over the past one year across days.
This is significant as it confirms trend continuation from the previous day as also a market which is very much in sync with it's global peers.
To expand the scope of the discussion, over the weekend, I intend to post some more stats on gap opens, opens out of prev ranges, gap closures, action post gap w.r.t the prev day's high and lows and some more.
Hi Kris and RM :
Thanks for the post, good insight into behavior of week day trading. Irrespective of whether trend is down or up, it is found most of gap down days is good for intra day longs and gap up days are good for intra day shorts (this is the 80:20 principle) and your findings complies the 80:20% principle once again very closely.
Another analysis that you will find interesting is the daily ATR levels for a period of 10 years and it clearly reveals which position is good to take home. Thanks once again and regards :). RM appreciate your efforts in assimilating data for some lucid analysis by Kris.
@tradesmyth, thanks for your input, it was educative.
Regards
Vinod
Thanks Vinod! Would it be possible for you to write a post on use of ATR in deciding carry forward trades?
regards
rm
Of course yes RM. Let me come back after my break refreshed in 2012 :)
Regards
Vinod
Post a Comment