Good question. Usually the 4 pm close of the market price. I don't like when there is a big movement in that 15 mins timeframe, and the market and the futures get out of sync. It isn't just 8 samples but my observaions from the last 2-3 years, nature abhores vacuum, the market abhores gaps...
I am not watching what's happening in US, but the ESTX and SMI just started looking like low-risk shorts. (They were bullish since the morning). Your rule may work today.
I've always been curious about the conditions that precede a gap fill. I took the time to go back over the past 10 years (until 1st quarter of 1998) Here's what I scanned for : 2 consecutive gap openings. This was simply a comparison of the open of the regular futures session (9:30ET)to the close (4:00pm) of the prvious session. I took the time to write code for Gap Ups and Gap downs. Separately. Gap had to be at least 1.25pts(>1.00) Here's the code I used for the Gap Downs. (remember, this is 2 (two) gaps down, one day after the other. I only looked at the SECOND consecutive gap down. ---------- [LegacyColorValue = true]; {file:strategy: gap001Dn} vars:gapfill(0),gapone(0),gaptwo(0); gaptwo=O-c[1]; if currentbar>2 then begin if c[1]-o>1 and c[2]>o[1] and c[2]>H[1] then value1=1; gapone= o[1]-c[2]; end; if value1=1 then begin If h>=(c[1]-.75) then gapfill=1 else gapfill=0; value1=0; value2=1; end; If value2=1 then begin Print(file("c:\101\gap001dn.txt"), date:0:0,",",open:0:2,",",c[1]:0:2,"," ,gapone:0:2,",",gaptwo:0:2,",",gapfill:0:0); value1=0; value2=0; end; ---------- I've attached an excel spreadsheet for this shit. Column titles should be sufficient to understand what's going on in the sheet. Here's the summary of two consecutive gaps down over the past 10 years using S&P e-mini 4:00pm close and 9:30am open. Also, I should point out that I took the liberty (do to personal experience) to give the ES 3 tics leaway. That means that I consider a gap to be filled if price comes within 3 tics (0.75 points) of the previous day's close. Overall (just talking about gaps down in this post) over the past 10 years there have been 36 times that the ES has experienced 2 consecutive gaps down. 25 of those times, the gap on the second day was filled (or within .75 of a fill). 25 out of 36 is 69% of the time. Pretty strong. In the attached excel sheet I included some of my own studies, for instance , I was curious about the size of the gaps. In other words if you added the two consecutive gaps together (the points lost from the close of yesterday (c[1]) to the open of today, was there a certain size that increased the likelihood of a fill. This is not something I though about until after I had run the data, but to get a % based level, I added both gaps together and divided byu the previous day's close. WHen the total of two days of down gaps was greater than (well, actually <-2%), the chances for a gap fill (OVERSOLD!!!) WAS GREAT. There were 11 times in the past 10 years that the 2 day's worth of gaps represented a loss of 2%, and 10 of those 11 saw the gap fill. 10/11 is 91% of the time. It's all in the spreadsheet. If I feel like it, I might analyze the Gaps Up tomorrow (This is a holiday afterall) DATE2nd in the first column of the spreadsheet is the date of the 2nd gap down. format is a little funky because it is TradeStation, 1070305 is 107-03-05 where: 107=2007, 03=March,05=5th day (of March)
Very interesting post and full credit to you for sharing your findings in such an open manner- imagine if this site was full of positive additions like this rather than people just shooting down or bigging up theories. Posting methodology and stats like this allows people to draw their own conclusions. However, in this case, your conclusions are pretty conclusive!!!!
It was pretty late last night when I wrote the post that included my first look at the 2 consecutive gaps down and I have to review the spreadsheet to make sure I didn't get a little confused. The data is right, I just have to double check the basic observations. I have been in the process of getting up early to observe and trade the European futures and I have been getting up at 2 or 3 in the morning (East coast); yesterday was one of those days, and I was pretty tired by the time I consolidated the spreadsheet and tried to make some obnservations.
Let me do some minor reiteration on the gap down study. (I'll consolidate all this later) RAW overview: I filtered prices for 2 consecutive gap down days (when the gap on the first day was never filled in RTH Regular Trading Hours being 9:30am to 4:00pm East Coast Time) RAW OVERVIEW: Since December of 1997, 2 consecutive gaps down have occurred 88 times, and 68 of those times, the gaps filled. (The gap created on the 2nd gap down day). 68 out of 88 is 77.3% of the time. (These are just the gap down day's, I haven't looked at the 2 consecutive gap up days, but I think this is strong evidence that Pekelo's 2nd gap rule has real substance (regardless of whether some anal retentive statistician wants to make a brouhaha about too small a sample group. Don't think of this study as a statistical tour de force, think of it as an historical perspective). "Overall (just talking about gaps down in this post) over the past 10 years there have been 36 times that the ES has experienced 2 consecutive gaps down" Pt of clarafication here: there have been 36 times when the total pts lost in the gaps represented at least 1% loss from close of 1st gap down day... Of those 36 occassions, 25 of those times, the gap on the second day was filled (or within .75 of a fill). 25 out of 36 is 69% of the time. Pretty strong." I plan to present a final report that includes all the salient points I have been able to uncover. I have included the data in the spreadsheet in case anyone out there wants to review dates and their own technical observations to share any additional filters that could mean $$$... and then share them -Anyway, I wanted to clarify, Over the past 10 years, there have been 36times that the points generated by gap on the 2nd consecutive gap down plus the pts generated by the gap of the first day, divided by the close of the 1st gap down day represented a loss of at least 1%. When that has been the case, 25 of those incidents (69% of the time) Gap on 2nd day fills. Now, when I filtered for times when the 2 day totals of gap points represented losses of 2% (or bigger losses), there were 11 occassions and in 10 of those events (90.9% of the time) the gap filled. EXPLANATION of gap today plus gap yesterday divided by close yesterday: These two sets of gap points totaled and divided by the close on the day of the first gap are presented in column "H" of the spreadsheet. This is what is ultimately presented in the column H: (((O[1]-C[2]) + (O-C[1]))/C[1]) *100 Why did I bother to do that?, Because I wanted to see if there was a threshold of "fear" or "pain" that boosted the chances for a gap fill. I needed a % based measure as opposed to simply pts. If I had thought about this before I had collected all the data I would have included the close of the day previous to the first gap down as a column in the data output, BUT, I didn't, so, I measured what I have, it's a little unorthodox, because a huge drop on the first day would give a smaller close (more likely to generate a big % change for the gaps), but what the hell, that is part of the first "wave" of fear, so maybe it carries validity of its own. Anyway, that's the way it went down for some of the measures that might be construed as unorthodox. I am in the process of consolidating the GapUp days to see if there is anything apparent in that data.
Vertigo, thanks for your evaluation, seems to spport my simple observation. About yesterday. YM futures closed the gap, the Dow stopped at +7 points. The Nasdaq cash also filled the gap, although the S&P and ES futures didn't, but tried. All in all I would say 2 out of 3 filling the gap isn't bad... Nevertheless all of them dropped in the first 10 mins of trading so if one shorted from the time of my post, there was a 20+ YM and 3+ ES points to be made....