Thursday, December 31, 2009

A Review Of The Modified Donchian Channel System

(I accidentally posted this before it was edited and complete, so check out the finished post.)

The Donchian Channel trading method was popularized by Richard Dennis as part of his Turtle Trading system. It could not be simpler. Just go long an N period high and then sell or sell short an N period low. Clearly the success of the Turtle traders proves the validity of this method. doesn't seem to work too well with stocks and stock indexes. In the book, Way of the Turtle by Curtis Faith, the system tests results are based on commodities, currencies and treasuries, but not stocks or stock indexes. (This is a great book for understanding the philosophy behind trend following systems.) I think there is a good reason for that. While I haven't run any tests myself on the original system, anyone who tries to trade it on stocks will quickly get frustrated with the long strings of losses and small profits with just enough big wins on occasion to make you think it's going to work. The obvious problem using the method with stocks led to the Turtle Soup method, which is essentially the same system in reverse, and has led many to claim that it doesn't work at all.

Of course, this is nothing new. I've heard it all before: moving averages don't work, the macd doesn't work, elliot wave doesn't work, trend following doesn't work, and on and on and on. An analogy might be helpful. Imagine someone who has no experience with golf, who has never seen a professional or successful golfer play, but perhaps has heard about the joys of playing golf from friends and family and wants to learn how to play. This person goes out and buys an expensive set of golf clubs and a couple of books, and then heads off to the driving range. How hard could it be? Just hit the little ball with the club. We all now how it's going to turn out - slices, blocks, hooks, shanks, poor distance and frustration. Even after expensive lessons there is likely to be little improvement. The novice soon concludes that it is impossible to consistently hit a long straight shot with a golf club. Whoever invented the golf club surely didn't know what they were doing. Except we know that isn't true. We have many great examples from professional golf that prove that one can hit great golf shots with a golf club. The problem is there are very few people who know how to teach other people how to use one even if they can use it well themselves. The same holds true with trading.

Using the Donchian Channels for stocks and stocks indexes requires some minor tweaking and a little discretion to make it work. I call this the Modified Donchian Channel System. The changes from the original system are based on a couple of observations. First, stocks don't trend as well as commodities and currencies. Second, stocks have an underlying bullish bias. And third, stocks tend to have more false breakouts. So the first modification is that we require CONFIRMATION on all sell and sell/short signals, and long signals that fail to CLOSE at new 20 day high. The second modification is that we do not go short on all sell signals, but only go SHORT when the stock or index is BELOW THE 200DEMA. Third, we use FAILURE SWINGS to help indentify turning points and avoid false signals. And finally, 20 day lows are often a great time to buy stocks.

Let's look at the Qs during 2008 with two channels on the chart. The channels are 20 days and 10 days with 1 day forward offet. The 20 day channels are for buy and sell signals, and the 10 day channels are to help identify failure swings. They should be used for exits only.

The first short signal occurs on 1/4/08. This is a valid sell signal because the previous 20 day low occurred on 11/12/07, there was no intervening 20 day high (closing basis) and the Qs were below the 200dema. On 1/4/08 the Qs closed below the 11/12/07 low CONFIRMING the short signal.

On 3/18/08 there was a failure swing. On 1/23/08 the Qs made a new 20 day low, and then 6 weeks (a long time) later on 3/10/08 there is a new 20 day closing low. There was another attempt with a new intraday 20 day low on 3/17/08, but on 3/18/08 the Qs closed above the highest close of the previous 12 days. So we exit short, but do not go long. This one is somewhat discretionary.

On 3/24/08, the Qs close at a new 20 day high. This is a long signal. After a downtrend with a long period of sideways consolidation and a failure swing, CONFIRMATION is not required.

The double top on 6/5/08 followed by a new 20 day closing low on 6/11/08 confirms a swing failure, and we exit long. This is not a short signal. On 6/24/08, the Qs make a new 20 day closing low below the 200dema CONFIRMING the sell signal and we go short.

On 8/6/08 the Qs make a new 20 day closing high and we exit short and go long. On 9/4/08 the Qs make a new 20 day closing low. On 9/5/08 there is another new closing low CONFIRMING the sell signal so we exit long and go short since the Qs are below the 200dema.

On 1/2/09 the Qs make a new 20 day closing high, so we exit short and go long.

Below is the same chart of the Qs for 2009.

On 1/24/09, there is a new 20 day closing low. We exit long and go short. I should've waited for another new closing low to CONFIRM the signal, but hey, I'm not perfect. On 2/6/09 there is a 20 day closing high, but it fails to close above the 1/6/09 high so we need to wait for CONFIRMATION, which does not come as the Qs do not close above the 2/6/09 high on 2/9/09 and then head down to the 3/9/09 low, which is not a new bear market low, but a retest of the 11/21/09 low.

On 3/17/09, there is a new 10 day closing high, which would have been the best place to exit short due to a failure swing. I missed it.

On 3/18/09, the Qs make a new 20 day high, but close poorly making it a good idea to wait for CONFIRMATION. CONFIRMATION comes on 3/23/09 and we have been long ever since.

I realize these rules do not permit a completely mechanical approach, at least not easily. However, they can be implemented successfully with a little experience. You only need to use common sense, wait for CONFIRMATION when prudent, and learn to identify swing failures.

The above examples show that one can be successful even with a few mistakes. Let's get real folks. We are going to make mistakes. To pretend that we aren't is to deceive ourselves. I hope this helps.

No comments: