*For each market, the first colimm shows the percentage of total cases and the second colimin gives the percentage that Mondays close was in the X direction.

{SYSTEM: Weekend Study I

ICopyright 1985-1998. PJ Kaufman, All rights reserved,I

(Update weekend study in Trading Systems and MetHodsl

vars: dDw(D). ix(0). nx(0), error(D), Mon(O), 1 (0), cesesiO),

div(O), pc(0), prevfrKO), savedir(O), start(O), change(O), TFopenfO), TFclose(O), TFˆxt(0), FFoper(D), FFclose(O), FFext(O), Fr4open(0), FrqclosetO), Frqext(O); arrays: week[16](C), Dweek[l6](0), xweek[16](0), pattern[5](0), cprice[5](0);

dow - @deyofweeK(date);

if currentbar - 1 then begin

weekCl] - 1011.; week[2] - 10110; weeK[3] - 10101; [1] - 10100;

v-eek[5] = 10011; week[6] - 10010; week[7] = 10001; week[8] = 10000;

v-eek[9] = mil; v»eek[10] - 11110; weekFll] - IIIOI; week[12] = 11100;

week[13] = 11011: week[14] - 11010; week[I5] = 11001; week[I6] - 11000;

end;

I Monday!

if dow = 1 then begin (Process weekend petternl

if Start = 1 then begin

(Thiirsdey to Friday direction!

change - closeCl] - close[2];

if Copen - closetll) * change > 0 then TFopen = TFopen + 1;

if (close - closetn) * change > 0 then TFclose - TFclnse + 1;

if (change > 0 and high > closeFl]) or (change < 0 and low < close[IJ)

then

TFext = TFext + 1;

(Friday to Friday din

change - closeCl] - prevfri;

if (oper - close[l]) • change > 0 then FFopen - FFopen + 1; il (close - CloseCl]) * change > 0 then FFclose - FFclose + i if (change > 0 and bigh > closeCl]) or (change < 0 and low < closed]) then

FFext - FFext + 1;

end;

erd; pattern[l] - ij weekly - d;

if error = 0 then begin icreate a weekly pattern!

weekly = 10000 + pettern[2]*1000 + pattern!3]-100 + patte .;:-:i -

pettern[5];

end; start - 1; end;

if lastcalcdate - date then begin

print ("thursday to friday*, cases:fl:0, tfopen;4:0, tfclose:4:d, re.t--i,:.

tfopen10d/cases:4:0, tfclose«100/cases:4:0,

ifexf"100/ceses;4:0); print ("friday to friday, cases:4:c, ffopen:4:0, ffcl Dse Ij ;d, ffext:--.:.

ffopen*ld0/cases;4:0, ffclose*10c/case5:4:0,

ffext*ld0/cases:4:0); print ("most frequent direction", ceses:4:0, frqopen:4:0, frqc]ose;4:c.

frqext:4:0, frqopen*!00/cases:4;0, frqcloseloo/cases: : ,

frqext*ldo/cases:4;0), print ("all patterns:); for i* - 1 to 16 begin

print week[ix]:6:0, xweek[ix] + oweek[ix];4:0,

(xweek[ix] + oweek[i x] )*100,cases:4:1. .

xweek[ix]*lod/(*week[ix]+oweek[ix] :i:C-:

end;

end;

A Comment on Testing and Holidajs

The weekday and weekend tests did not include any weeks in which there was a holiday or long holiday weekends. It would have been perfectly reasonable to view the opening of the first day of the following week as a "Monday" even though it occurred on a Tuesday following a Monday holiday. In fact, those special cases may have proved very consistent Simuarly, a holiday on Friday could have been worked out to fit the weekday pattems and allow the Monday direction to be added into the total results. These tests shown here only used complete 5-day trading weeks.

A well-known work that studies price movement prior to holidajs is by MerriU, whose results are based on the stock market and demonstrate a strong buUish tendencj before a holiday with a weak day immediately after. Remembering the buUish bias of the stock market (about 54"o of all dajs were higher ftom 189* to Januarj 1964), Merrills results are shown in Table 15-9. In relation to trading, this would indicate the possibility of a sharp ftending move prior to or throughout a holiday season.

TABLE 15-9 Merrills Holiday Results PeriwfTesterf Holiday or Holiday Period % Ufward Mores

1897-1964 Day prior to all holida/s 67.9%

1897-1964 Day after all holidiys 50.8

1897-1964 Thanksvinjio New Year 74

1897-1964 July 4chio Labor Day 69

1931-1965 Before Christmas 74

1931-1965 Before NewYear 75

The Januarj Effect

Another bias m the pattem of prices may be seen in the action of the stock market during the month of Januarj. There are many investors who are not as anxious to trade in and out of the maiket as professional