Well, I have always been a skeptic of candlestick patterns, but I thought It would be best to test them. So I put together a bunch of rules that are meant to emulate the “high reliability” bullish patterns as found on:
http://www.candlesticker.com/Bullish.asp
I tried to convert them into rules for P123, in order to test:
Piercing Line
close(1) < open(1) AND open(0) < close(1) AND 2*close(0) > close(1) + open(1)
Kicking
close(1) < open(1) AND open(0) > open(1) AND close(0) > open(0)
Abandoned Baby
close(2) < open(2) AND hi(1) < low(2) AND low(0) > hi(1) AND close(0) > open(0)
Morning doji star
close(2) < open(2) AND open(1) < close(2) AND open(0) > close(1) AND open(1) > 0.99close(1) AND open(1) < 1.01close(1) AND 2*close(0) > open(2)+close(2)
Morning star
close(2) < open(2) AND open(1) < close(2) AND open(1) > 0.95close(1) AND open(1) < 1.05close(1) AND 2*close(0) > open(2)+close(2)
Three inside up
close(2) < open(2) AND close(1) > open(1) AND open(1) > close(2) AND close(1) < open(2) AND close(0) > close(1)
Three outside up
close(2) < open(2) AND open(1) < close(2) AND close(1) > open(2) AND close(0) > close(1)
T
hree white soldiers
close(2) > open(2) AND open(1) < close(2) AND close(1) > close(2) AND open(0) < close(1) AND close(0) > close(1)
So I wanted to see if they had any effect. I set a rule for “nodays > 10” as the sell rule, to allow a 10 day holding period. I then ran sims with each of the above rules, to see if it would outperfom a “control” case, with no rules at all with a non technical ranking system.
None of the above patterns appear to work, i.e. outperform the control. If you want to run for yourself just copy the following sim, and uncheck the rules as required:
http://www.portfolio123.com/port_summary.jsp?portid=261882
I am hoping to invite some sort of “peer review”, have I got this wrong? I had the impression that Stitts may think candlesticks did work, I would be very interested in being proved wrong!
In other words and joking aside, they may have looked pretty darn good when there were scant data, but not so great in the data immersed world in which we live. Who knows? I wouldn’t discount anything.
