I’m undertaking a 1000-day reinvention project, blogging here daily to track my progress. In Monday Money, I write about money management.
Trading based on feelings and convictions doesn’t work, I’ve learned through expensive lessons. Trading based on trend following, attention to charts (which can tell you when a trend is reaching exhaustion or has reversed), diversification (because you can’t predict which sectors will do well), and broad market sentiment does.
I came across a stock market indicator approach that I hadn’t heard of before over the weekend: the Donchian channel (DC). This can help you identify new bullish or bearish trends, breaking out from previous channels (areas of consolidation, where an asset price is staying confined to a particular range).
A DC uses the highest high and the lowest low of the last n periods (e.g., last six months, last 50 days, etc.) These bullish and bearish extremes can indicate a breakout up or down, and suggest when you should buy, sell, go short, or exit a short.
Portfolio manager Paul Mulvaney, who claims 20% per year returns for more than 26 years of trading, reportedly uses Donchian channel breakouts to identify new long and short entries in his commodity trading approach. A reverse engineering of his rules suggests he uses six-month lookbacks.
Mulvaney doesn’t apparently use a simple DC-based indicator for exiting long or short trades, but uses probability estimates to place stop losses and reposition them daily “in accordance with a volatility analysis.”
Turtle trading using Donchian channels
As I was learning about Donchian channels, I discovered an interesting commodity trading experiment that was completed in 1983. Commodity traders Richard Dennis and William Eckhardt wanted to see if great traders are born or can be trained. A set of complete beginner traders were given a trading rulebook based on Donchian channel breakouts and tested to see if they could do as well as experienced traders. The beginner traders successfully made money, generating a reported 150 million dollars, based on an initial kitty of about $1 million for each of 23 traders.
The rules the traders in the experiment followed were:
- Focus only on highly liquid futures markets with tight spreads and deep orders books, to ensure smooth execution, reduced slippage (losing money because of the difference between bid/ask spreads), and clean price movements.
- Use position sizing to manage risk. Limit initial risk to 1% of capital using the formula position size = (1% of capital) / (2 x ATR) where ATR is the average true range, or the difference between the most recent high and low for a period (sometimes 14 days).
- Enter positions when an asset’s price exceeds the high of the past 20 days (the 20-day Donchian channel). Go short an asset when the price falls below the low of the last 20 days.
- Set an initial stop loss at 2N below the entry point for a long position or 2N above for a short position. This can help cut losses quickly if the market doesn’t move as expected.
- Use trailing stops to protect gains. Adjust stops to a 10-day low for long positions or a 10-day high for short positions, locking in profits while still allowing room for the returns to grow.
- Use pyramiding to increase the position in increments as the market moves favorably. Add another unit of the trade as the market moves every 0.5N in the right direction, up to four units total. There were other limitations on risk based on correlated markets and how much long or short the traders were (12 units maximum long or short).
Can it work now?
A backtest of some simplified turtle trading rules found that it didn’t do well over the past 20 years, over 4322 trades. It had a winning rate of 36.83%, an annual return of -0.38%, and a max drawdown of more than 95%!
Rayner Teo, who ran the backtest, also tried a modified version that increased the number of markets to trade, reduced risk per trade, and increased the length of the breakout. In this backtest, the winning rate was 41%, annual return 32%, and maximum drawdown -42%. In this case, the entry was a breakout above the 200-day high (instead of 20-day), a stop loss of 2 ATR from the entry price, a trailing stop loss of the 10 day low, and risk management reduced to 1% of the account from 2%. Of course, reverse the numbers for short trades.
Teo offers the following three lessons from his analysis:
- Understand the logic and concept behind your trading strategy. Based on this you can develop multiple trading strategies and diversify your risk. Understand it well enough that you don’t abandon it when a drawdown comes.
- Manage your risk. In a modified backtest of the modified turtle strategy he analyzed he increased max risk to 4%. In this case he saw an annual return of 76% but a maximum drawdown of 96%. You can’t recover from that! You blew up!
- Adapt to changing market conditions. I don’t see this as an implication of his analysis but rather a meta-principle that a particular trading strategy won’t work all the time. I’ve been trying the options wheel with a lot of success lately but if we get into a grinding bear market, it won’t work well.
How can I use this?
I’m thinking about how to work this into my trading. I want to enter and exit trades based on indicators not on feelings. The DC has the benefit of being simple. But it needs to be used alongside key practices around diversification, position sizing/risk management, and which trades to do in which situation. That’s an analysis for another day.