Sunday, October 3, 2010
Thus is a study I just completed. Basically I plotted
1. white candle days minus black candle days (Shown in subchart 4 with a StdDev)
2. close > previous close minus close < previous close (Shown in subchart 5 with a StdDev)
3. The difference between these two plots. (Shown in subchart 1 without a StdDev)
I also plotted these times volume. (subcharts 2 and 3)
Nothing really popped out at me but closer inspection of subchart 1 does seem to give an early tell to change.
It's not a strong indicator, but it's not worthless. This is the type of pains taking research I do everyday, and it is how I've come to invent so many of my indicators.
I'm currently working on a new project that could take months. (it's only work if you don't enjoy it.. ;- )
Here's the idea:
Confirmation is one of the most important parts of candlesticks.
I've heard people say things like a Doji is indication of a peak or valley.
Truth is a Doji on it's own stands about a 50% chance of being a peak or valley.
While I'm impressed with Bulkowski's work in the Encyclopedia of Candlestick Charts, it seems to be lacking data involving confirmed conditions. My research shows that 5 days after a "Doji Turn" (a Peak or Valley with lower highs or higher lows to either side), the market has a 60.5% chance of actually having reversed over that 5 days. If the Doji is proceeded by a Long candle the odds improve to 65.5%
There are hundreds of conditions to test, for example - are Valley Doji better than Peak Doji or what happens if the confirming candlestick is long or the Doji has higher than average volume?
This research is time consuming, and requires a lot of careful coding, and thinking about how the condition should be constructed. Keep in mind that coding filters for near certain conditions will be useless if the condition is only found once in five years. Currently I'm debating how to define a reversal, the trend, and how to define a Bear market or a Bull market. As you can see this is a complex study.
I've only begun this research using data which includes Open, High, Low, Close, and Volume back to 1950. It's been a while since I've written spreadsheets, so I'm enjoying the learning curve again, but it's returning quickly. I'm using Open Office because I feel it is more universal, and available to everyone. It also seems to work better and quicker than Excel.
I'm wondering if others might wish to join me in a collaborative effort to unravel the hidden truth about candlesticks and candlestick patterns.
This Blog has only informational, and educational purpose, and does not represent a proposal for buying or selling contracts, equities or currency.
Posted by Robert Campbell at 10:40