d1d2.us | blog | contact | d1d2.csv

Blog

2015-11-12

Way back in 2013 I conducted a study with a simple question:

Are some candlestick-patterns accurately bullish or bearish?

If you are new to candlesticks you can learn about them on Wikipedia:

https://en.wikipedia.org/wiki/Candlestick_chart

I designed the study to capture all 3-day candlestick-patterns of the S&P 500 between 1983 and 2013 which gave me a list of about 30 x 252 observations.

Next I matched each pattern-observation with its 4th-day-gain (or loss).

Then I used a long switch-case statement to categorize the observations.

Next I sorted each category by average 4th-day-gain.

I found that almost all of the patterns corresponded to a gain near 0.038% which is also the average gain for all the observations.

So, this convinced me that candlestick-patterns are not very useful.

I did find, however, evidence of 'mean-reversion' in the gains.

For example if I collect all days which have a delta between -3% and -1% it is easy to calculate the average of the next day to be 0.57% which is 50% higher than the 0.038% gain for all observations.

And thus the idea of d1d2.us was born.

This site, d1d2.us, allows me to compare today's gain to past one-day gains.

I designed the site so today's gain will land in a bin of a bright green histogram.

If that bin is higher than the median, I see that as a bullish signal.

Each bin is 1/3% wide.

For example if the S&P 500 is trading at 0.45% below the previous close, that puts me in what I call the minus two bin.

I can see from the histogram that this bin has a median gain which is higher than average.

And so, if the S&P 500 is trading at 0.45% below the previous close, I consider that a bullish signal.


d1d2.us | blog | contact | d1d2.csv