Candlestick Pattern – Doji

My young friend in Prague put up an interesting chart the other day: QQQ and the doji pattern:

A doji is classified as a “reversal” pattern, but that doesn’t mean that prices should reverse. It is more correctly viewed as simply the end of a trend. The market has been going in some direction, in this case down, and then one day prices finish roughly where they started. It is an indication that the bears might be losing control after knocking the index down 4% since last week. It does not necessarily mean we can expect a bullish upswing.

I have tested candlestick patterns so many times, and they have all been basically worthless in terms of any predictive value. However, I have never tested it on indices, only on stocks. I always use candlestick pattern when looking at charts, but that’s just because it’s more visually appealing to me (and perhaps easier for me to find testable ideas). But I wanted to have a look at the doji pattern and much to my surprise the results are quite good.

But first, let’s start with a explanation of the pattern:

I have candlestick book written by Gregory L. Morris in my bookshelf called Candlestick Charting Explained. That’s a book I bought back in 1998. I looked up “doji” and it’s a pattern that is classified as “spinning top”. Morris writes:

Spinning tops are candlestick lines that have small bodies with upper and lower shadows that are of greater length than the body’s length. This represents indecision between the bulls and the bears….The small body relative to the shadows is what makes the spinning top.

Morris then goes on explaining types of dojis: doji, long-legged doji, gravestone doji, dragonfly doji, four price doji, stars, and paper umbrella. Quite exotic!

Here is an example of a doji:

As you can see, the “body” is very small and the “shadows” are long. And the doji made a perfect bottom before it rose upwards.

 

Let’s look at it more systematically and test this on SPY, the ETF for S&P 500.

Metastock has a built-in pattern recognizer for candlesticks. I simply used that one and ran a test on dojis for SPY.  The only condition I used is that at least one day of the last two days must be a down day (including the doji). Exit is on the close after two consecutive up days. Here is the result:

 

45 trades and 36 winners. The average gain is a respectable 0.76% per trade (34.2% in total. The graph shows the return with just 50% allocated capital). The only down year is 2008.  The doji is supposed to be a reversal pattern. What happens if we only go long if the close is below a 10 day moving average?

 

The number of trades is reduced to only 25, but 22 are winners. Average gain is 1,39% per trade, quite respectable. So what causes this good result? Luck? Good exit? I think the main reason is that the market is a somewhat “oversold” and we sell on a bounce. The doji might have some value, but I’m not sure.

What if we turn it upside down and go short? Here is the result vive versa compared to long (and the close of entry must be bigger than the 10 day average of the close):

 

 

32 trades, 19 winners and average gain of ,57%. Pretty good for a short strategy! In general, short is always a lot more trickier than long strategies. Only down years are 2006 and 2012. My experience has taught me that the market drops a lot faster than it rises. So let’s put in a target on short of 1%:

 

A profit target does not get any improvement, it gets worse.

So to sum up, this strategy might be tradeable, but only together with other alternative ones.