Do stock traders front run three-day weekends?

The Day before 3 Day Weekends

In my formulative years, the trading day before any three day weekend had a distinct bullish bias, up over 82.6% of the time from 1950-1975. This was eventually documented by students of the calendar and I first read about it in the mid 80s and as I have mentioned often, one of my Uncle Yogi Berra favorite quotes, 'Once something is obvious, it ain't obvious no more'. Over the last 10 years, the day before all three day weekends is only 34-30 for a median gain of 0.05%. And once it comes out in Print, you would almost do as well to fade it, especially if the market shows signs of frontrunning the trend. For example, on those 8 occasions in the last ten years, where the 2nd day before a three day weekend was up at least 1%, the S&P was up only twice on the actual day before a three day weekend for a median loss of 0.90%. The S&P was up 1.3% today. Day -1 in the table represents the day before the three day weekend and Day +1, the day after.

The Day Before a 3 Day Weekend
When Previous Day > 1.0%

DATE DAY-2 DAY-1 DAY+1
20020118 1.00 -0.99 -0.73
20020524 1.02 -1.21 -0.86
20030703 1.17 -0.81 1.90
20040903 1.12 -0.42 0.69
20060526 1.11 0.57 -1.59
20080829 1.48 -1.37 -0.41
20090409 1.18 3.81 0.25
20100528 3.29 -1.24 -1.72

#UP-DN = 2- 6 3- 5
AVG%CHG= -0.21 -0.31
MED%CHG= -0.90 -0.57
1%MOVES= 1- 3 1- 2

The Bulls will argue that today's gains were more about some Gonzo earnings announcements from IBM, Intel and Apple after hours than frontrunning of seasonal trends. I would like to believe they are correct and the fact that Globex SPs are up 6 points in after hours trading doesn't hurt their case.

The Last PTA6 Signal

Those of you who have followed my work over the last 12 months have a sense of where I am going with my tape studies. After a 17 year absence, there was a reappearance of three different very important ten day breadth thrust signals in 2009. I spent December of 2009, writing a thrust code to reevaluate the utility of breadth thrust signals and attempted to rectify the silent decade via evaluation of thrust with other tape measures and time intervals. In January of 2010, I documented my results, came up with a hook (PTA) for a title and passed it along to MTA for their appraisal. This was the beginning of the most interesting year of this Bean counters career, at least from a research perspective.

The original code was a concept code, written to demonstrate an approach to thrust modeling. I had finished version 3 by the time the 2010 Charles Dow deal was announced in May of 2010 and I have written 6 3/4 different versions of the code. For quants in the audience, who get into this kind of stuff, it is important to understand that all 7 versions of the code, with the exception of 6, are 'closed form' solutions. That is, I have already done all the previous background research and instructed the daily PTA code what it is to scan for.

Version Six is the ultimate goal, a code which searches each day, with no input from the developer other than to look at the 7 different defined measures of tape activity across N different time intervals in search of those extremely rare 4 standard deviation level signals, bullish or bearish, which have historically driven the market for months. The PTA5 and PTA7 ratings (0-100), which I consider my best closed form approaches, are listed each day in my commentary. I haven't quantified my cumulative rating system for PTA6, but let us revisit the last PTA6 signal of March 21 of this year.

Under my current PTA6 constraints, below is the only PTA6 tape signal for 2011, albeit, a nice one on March 21 when an ADT3 of 82.40 was registered. The matches are weighted, so that the top 15 get 75% of the overall weighting. Note in this case, the top 22 matches, or UDT3s above 81.3, were higher one year later. If I were simply trying to convey a bullish message, I could have excluded the bottom 9 from the table.

ADT3 ANALYSIS OF 79.46 to 83.46
FOR THE ADT3 OF 82.40 OBSERVED ON MAR 23,2011
ADT3 =100*ADV/(ADV+DEC) OVER 3 DAYS

# DATE RANK LEVEL 1MT 3MT 6MT 12MT
0 20110321 0 82.40 ? ? ? ?
1 19780417 1 82.22 5.45 3.53 7.21 7.19
2 20090716 2 82.62 6.73 15.62 20.76 13.20
3 19840802 3 82.64 5.50 5.97 13.06 21.20
4 19821007 4 82.15 10.37 12.76 17.83 32.61
5 19750103 5 82.15 10.06 15.27 33.45 28.55
6 19700529 6 82.75 -4.78 6.94 12.25 30.15
7 19870106 7 82.89 10.78 19.45 20.63 2.42
8 19760106 8 81.89 6.34 10.51 10.70 12.28
9 20050518 9 81.67 2.65 2.82 5.29 6.43
10 19871209 10 81.61 1.89 12.63 13.11 15.97
11 19790104 11 81.50 0.93 4.13 3.56 8.05
12 20021015 12 81.46 3.24 4.19 1.08 18.78
13 20030317 13 81.46 3.57 17.25 18.91 30.25
14 20040818 14 83.46 3.05 8.07 9.72 11.31
15 20081126 15 80.94 -1.68 -15.19 2.55 25.11
16 19911226 16 80.79 2.63 0.75 -0.34 8.63
17 20030902 17 80.78 -0.17 4.37 12.44 9.42
18 19901022 18 80.60 0.40 4.30 21.03 23.21
19 19741011 19 80.57 5.64 2.07 18.33 23.99
20 20091109 20 80.55 0.26 -2.06 1.63 11.01
21 19960802 21 80.43 -1.58 6.23 18.67 42.97
22 20080226 22 80.21 -2.91 -0.39 -7.95 -45.50
23 19970430 23 80.19 5.86 18.84 12.77 38.75
24 19801113 24 80.09 -5.32 -6.97 -4.35 -10.86
25 20050225 25 79.96 -3.30 -1.76 0.08 6.44
26 20100218 26 79.80 5.34 1.27 -1.14 21.35
27 19791112 27 79.58 3.87 13.90 1.23 30.03
28 19870526 28 79.54 6.24 15.72 -15.57 -11.93
29 19971231 29 79.53 1.02 13.53 16.84 26.67
30 19860813 30 79.46 -6.11 -1.08 13.85 36.22

#UP-DN = 22- 8 24- 6 25- 5 27- 3
AVG%CHG= 3.71 7.44 11.78 16.32
MED%CHG= 2.85 5.17 11.48 17.37

A very similar ADT3 of 82.65 was observed on Dec 3, 2011 which is still in play as well. Although, the tendency of the market is to consolidate a day of gains like todays, it will be interesting to see if the market still has enough testosterone to pop up on PTA6's screen over the next couple of days. We had 438 Advances on the S&P today and Globex trading (+6) is encouraging.

Of course tape activity is not the only variable in making investment decisions, but technicians enjoy commenting that they are convinced the tape reflects a large majority of what is going on elsewhere. I believe in keeping one eye on the fundamentals as well, but appreciate the technician's position in this regard.

G. Wayne Whaley has been executive vice-president for Witter & Lester since 1993 and principal since 1998. He holds a B. S. degree in the Science of Mathematics from Jacksonville State University and a M.S. degree in Operations Research (Applied Math) from Georgia Technological Institute. He has been registered with the Advisor as an Associated Person since July 14, 1995, and as a Principal since, March 20, 1998. More info: http://witterlester.com.

Comments
comments powered by Disqus