Abstract

In this tutorial we will take a close look at the Dynamic Breakout II strategy based on the book Building Winning Trading Systems.

First we decide the look-back period based on the change rate of volatility, then we make trading decisions based on the highest high and lowest low from the look back period as well as a Bollinger Bands indicator. It is an auto adaptive trading system that can adjust its buy and sell rules depending on the performance of these rules in the past. In addition to Forex markets, it is widely used in Future and Equity markets. You can refer to this video to learn more about dynamic break out II.

The original Dynamic Break Out system was developed by George Pruitt for Futures Magazine in 1996. The logic behind the dynamic breakout system is that the volatility component changes the lookback period, then the enter and exit points are decided by the highest high and lowest low price over the lookback period.  The newer version of the Dynamic Break Out is just like the original, except we introduce the Bollinger Band and adjust the number of look back days using the market volatility, so different market conditions perform better with different parameters. In addition, the stop loss signal is fixed in version one, but in version two the liquidate point is based on the moving average indicator and the length of moving average is dynamically changed with the look-back period.

We backtested the strategy on EURUSD and GBPUSD over 6 years period.  The result suggests a drawdown of 20% and the strategy caught the market turning points. It is especially profitable in a trending market.

Method

Step 1: Determine the look back periods

The lookback period is the number of bars back from the most recent bar that the price or indicator looks at to make the momentum calculations. To start the look back period is set to 20 days to determine its buy and sell levels. We change the number of look back days in proportion to changes in market volatility. Through this method the number of look back days changes on a daily basis. At the end of each day, the current market volatility is calculated by the standard deviation of the past 30 day's closing prices.

close = self.History(self.syl, 31, Resolution.Daily)['close']
todayvol = np.std(close[1:self.numdays+1])
yesterdayvol = np.std(close[0:self.numdays])
deltavol = (todayvol - yesterdayvol) / todayvol
self.numdays = round(self.numdays * (1 + deltavol)) # the number of days must be integer

Though the look back days are dynamic, it needs to be restricted within an acceptable range of 20 to 60.

Step 2: Choose the algorithm buy/sell point

For a buy setup, the close price of the previous day must be above the upper Bollinger Band. In addition the ask price must be above the highest high of the most recent \(N\) days. Where \(N\) is the look back days from Step-1. For a sell setup, the close price of previous day must be below the lower Bollinger Band and the ask price must be below the lowest low of the most recent \(N\) days. The length of the Bollinger Band calculation is the same number of look back days that is generated by Step-1.

Bollinger Band is a popular technical indicator. \(k\) is a constant. Here we choose \(k\)=2.

\[ Upper\ Band = \mu + k\times \sigma \]

\[ Lower\ Band = \mu - k\times \sigma \]

where \(\mu\) is the moving average and \(\sigma\) is the standard deviation. QuantConnect provides more than 100 technical indicators for you to use in your algorithm. These are provided as class objects in Python. A full list of the indicators and their properties can be found in the Supported Indicators page.

self.bolband = self.BB(self.syl,self.numdays,decimal.Decimal(2),MovingAverageType.Exponential,Resolution.Daily)
self.upband = self.bolband.UpperBand
self.lowband = self.bolband.LowerBand

Step 3: Choose the algorithm liquidation point

The exit signal for an existing holding is determined by calculating a simple moving average of closing prices for the past look back days. That is to say, we liquidate a long position if the current price is lower than the moving average of the close price over the look back period, and vice versa for selling a short position.

self.buypoint = max(self.high)
self.sellpoint = min(self.low)
historyclose = self.History(self.syl, self.numdays, Resolution.Daily)['close']
self.longLiqPoint = np.mean(historyclose)
self.shortLiqPoint = np.mean(historyclose)
self.yesterdayclose = historyclose.iloc[-1]

Conclusion

For six years backtesting of EURUSD, the overall statistics show an annual rate of return of 2.3% and with a Sharpe Ratio of 0.31. EURUSD has a significant uptrend from 2010 to 2012. This momentum strategy outperforms the market and seems to be profitable from 2010 to 2014.  The maximum drawdown occurs in May 2015 to December 2015 and is roughly 14%. From our results we find the strategy works best in an trending forex market.

In contrast, GBPUSD is pretty volatile during the tested period from 2010 to 2016. Our testing demonstrated a negative annual rate of return with a drawdown of approximately 19%. When the volatility decreases, the price tends to continue following the current trend. Volatility causes the the look back days to decrease when computing the bollinger bands, making it easier to enter a trade.  If the market volatility increases we increase the look back days in order to filter the fake signals, making it harder to enter a trade.

Here we use the standard deviation of price as a measure of market volatility. To improve the model we could choose other measures of volatility like standard deviation of logarithm return series or other stochastic volatility measures.


Reference

  1. George Pruitt, John R. Hill, Michael Russak (September 2012). Building Winning Trading Systems, page 126,  Online Copy
  2. Robert C. Miner(October 20, 2008). High Probability Trading Strategies: Entry to Exit Tactics for the Forex, Futures, and Stock Markets, page 37,  Online Copy

Author