Historical Performance of Trend Following Post-Drawdown
August marked a local low for our Trend Following index, reaching about two thirds of the max historical drawdown. We were commenting on the fact that, historically, drawdowns have typically proved a good time to invest or start trading a trend following strategy.
This prompted us to run a test to check the historical results on the index. We decided to investigate the historical performance immediately following the first day of a large drawdown (2/3 of MaxDD reached).
The backtest for the index starts on January 2000. We let the backtest run for 5 years, to establish an initial baseline maximum drawdown (MaxDD). We then picked all occurrences thereafter where the drawdown reached
two thirds of the MaxDD at the time (limiting the number of occurrences to one per year to limit overlap). Measuring and plotting the performance of the index past these occurrences gives an indication of the historical post-drawdown average performance.
For context, below is a chart of the equity curve and drawdowns for the full duration of backtest and live test (performance is hypothetical):
From 2005 onwards, there were 5 occurrences/years during which the drawdown reached the MaxDD two-third level. Despite this, the index still returned +435% (or +15.5% annualized), as drawdowns are an integral part of a successful trend following strategy. These are the dates for these drawdowns:
- April 2005
- May 2006
- March 2008
- January 2010
- May 2011
Below is the chart showing the equity curve post-drawdowns, for all these instances, plotted for a duration of 3 years each. A composite curve is added to derive an average tendency. Performance is hypothetical.
In some cases the drawdown extends for a period of time but most of the post-drawdown performance is positive (in 80% of cases), as illustrated by the composite curve, nearly doubling in value after 3 years.
Below are the average performances at different time horizons:
Note also that the index was launched in live monitoring (post-backtest) at high levels of drawdown and has returned over 34% since. It will be interesting to monitor the performance of the index over the next few months.
The index performance is simulated using Trading Blox and CSI data (back-adjusted contracts rolling on Open Interest). The performance of the index is directly derived from the performance of a Trading Blox simulation suite composed of each system component as a system part of that suite.
The simulation uses realistic trading friction parameters (slippage, commissions, interest as detailed aside).
For more information on the index (systems and portfolio used and other details), please check our latest State of Trend Following report.
Trade Friction parameters
|Slippage||5% of ATR|
|Commisions||$20 per contract|
|Slippage on rolls||Yes|
|Roll slippage||5% of ATR|
The test is set-up with an arbitrary starting capital of 1B, starting in 2000. As the test is intended to represent an hypothetical index, no liquidity/volume constraints are enforced, making the results less dependent on actual simulation capital.
Profits are compounded (assumed to be reinvested).
The purchase or sale price for each trade that generated the hypothetical results is based either on 1) open price, the day after the Buy or Sell signal for the Moving Average-based systems or 2) stop level set by the relevant indicator for the Bollinger or Donchian systems. The actual simulated fill price is obtained by calculating a slippage factor, which is added to (or subtracted from) the theoretical entry price. For a long entry, the slippage factor is calculated by measuring the range from the theoretical entry price to the day’s highest price, and multiplying that amount by the Slippage Percent. (For short
entries, the slippage factor is calculated by measuring the range from the theoretical entry price to the low). The slippage factor is then added to, or subtracted from the theoretical entry price, to obtain the simulated fill price.