blog_fxsystemdevelopment

Many traders nowadays say that “classic trend-following is dead”, “High frequency trading is the way to go”, “simple methods don’t work because everyone would be rich”.  Is this true?

Recently FXRenew published a piece on random entries , and the conclusion of the article was that random entries can still work in today’s markets, despite the fact that the long lasting trends in currencies and commodities seen in the 80’s and 90’s are no longer present. The key was actually filtering out the Market Type (Trend or Range) because we demonstrated how it was the overriding variable in our tests. It’s inefficient to fade strength or buy weakness in a trend; it’s inefficient to buy strength and sell weakness in a range.

In the second piece of research we did , we explored volatility cycles and how there is in fact a link between volatility and trends.

In this article, we are going to verify these claims and see how they hold up to inspection. By using simple mean reverting and continuation algorithms, we will filter market states. Furthermore, we will dust off an old trend-following approach used by the Turtle Traders, explore it’s logic and see if old-school trend-following still applies today.

Continuation vs. Mean Reversion

We previously stated that mean reversion is the normal state of equity markets whilst continuation (or trendiness) is the normal state of currency & commodity markets. To prove whether this is in fact true, we created simple algorithms.

  1. Continuation: if today’s close > yesterday’s close, then buy on tomorrow’s open and close at End of Day.
  2. Mean Reversion: if today’s close < yesterday’s close, then sell on tomorrow’s open and close at End of Day.

mr-euro

Mean Reversion algo on EurUsd from 1/1/2005 to 14/9/2016

mr-es

Mean Reversion algo on S&P 500 from 1/1/2005 to 14/9/2016

We could post many more equity curves but the results would be similar: it appears that capturing the real continuation dynamics and mean reversion dynamics is a bit more complex. However, even with this basic test it appears that there are cycles of mean reversion and cycles of continuation in the data.

One improvement on the basic continuation/mean reversion algo is to filter for volatility states. When volatility is low, expect mean reversion and when volatility is high, expect some continuation.

vol-cycles-eurusd

Volatility Cycles on EurUsd using the ATR Oscillator  (programmed by Craig Consulting)

However, volatility cycles do not completely explain the continuation/mean reversion dynamic. Things are slightly more complex than that.  However, at this point in our testing, Craig & I were starting to be tempted by the enemy: optimization. And optimization is the opposite of robust and sustainable trading, so we needed to rethink our approach.

The Basics of Continuation in Forex and Commodities

Most people fail at trend following because they focus on the wrong things. The get caught up with entry rules and indicators. They also focus on one asset class only (or worse: even on one single asset!). In our random entry article, we clearly demonstrated that a simple trailing stop is enough to keep traders in a trend, when there is a trend.

gbpusd-trend

GbpUsd Range vs Trend Phase

Instead, the rules of engagement are not important. Indicators are not important. Instead, the concept is what matters: enter every potential trend and never miss a breakout. Trend followers will never buy at the bottom or sell at the top. They will arrive late to the party but they will enjoy the ride. But the Key to professional trend following is diversification. Most Trend Following CTA programs (like Dunn, Millburn, Abrahams) take positions in 8-10 Agricultural Commodities (Cotton, Corn, Lumber, Sugar, etc.), 8-10 Non-agricultural commodities (Crude, Heating Oil, Copper, Gold, etc.), 8-10 Currencies (the major pairs plus EurGbp & EurJpy), 8-10 Equity futures (Dax, S&P 500, Nikkei, etc), 8-10 Rates futures (Bund, Shatz, Short Sterling, etc).

 

millburn

Millburn CTA performance since inception vs S&P total return and Altegris 40

Source http://www.managedfutures.com/program_performance.aspx?fundtype=&productId=18659

Now for the big questions: how can all these CTAs be consistently profitable, yet it’s so difficult for retail traders to trade trends?  The answer is diversification and capitalization.

If you look at a single market at any given time, there is a very high likelihood that no trend exists at the moment. That not only means that there are no profits for the trend-following strategies, but it also means that there are potentially constant losses as prices make false breakouts. Trend-following trading on a single instrument a futile exercise, mostly because it deviates from successful investment practices (like diversification) and leads to optimization (as you try to make something work). Any single instrument or even asset class can have very long periods where this approach simply does not.  Instead, with a well diversified strategy, you have a large basket of instruments to trade covering all major asset classes, making each single bet by itself almost insignificant to the overall performance.

Inside the Black Box: Building a consistent 4-Rule Trend-Following Algo

The core idea of trend following is extremely simple: buy breakouts from a given range (which is determined by the length of the lookback corridor). The method waits for directional volatility to build in one direction, and then gets on board.  Since trends are not the common state of the markets, expect to lose 60% of the time, and make sure winners are large enough to pay for the losses.

For our CTA-replication system, we are going to use simple Donchian Channels, named after the trader that made them famous, Richard Donchian, the father of the Turtle Traders. The core concept is to buy when price closes above the top channel and sell when price closes below the bottom channel (which will be N-Day Highs and Lows).

donchian-channels

Donchian Channels overlayed onto EurUsd and programmed by Craig Consulting

To demonstrate that the concept is valid and that the key to trend following is not the entry, here are various equity curves for our strategy, using different channel lengths.

euro-16-day-channel-no-sma

EurUsd 16-Day Donchian Breakout with simple trailing stop

euro-60-day-channel-no-sma

 

EurUsd 60 Day Donchian Breakout with simple trailing stop

euro-90-day-channel-no-sma

EurUsd 90 Day Donchian Breakout with simple trailing stop

As you can see, the equity curves are quite similar. Going for a longer breakout reduces the frequency of trades. But it’s the trailing stop that locks in the profits and allows the system to work. Make no mistake.

So if the length of the channel is not important, what are the important factors in Trend Following?

  1. Diversification: seek to incorporate multiple asset classes into your portfolio.
  2. Position sizing: seek to keep each position small and sized based on volatility, so the portfolio is never overly exposed to one trade.
  3. Time horizon: which trends are you seeking to capture? Perhaps combining multiple time-frames can help to smooth the equity curve.
  4. Capitalization: evidently, managing a professional diversified trend following strategy requires a decent sized trading account. Fortunately, brokers nowadays are becoming one-stop shops, giving retail traders access to multiple instruments at once.

So following the empirical rationale (not verified extensively in our research at this time) that:

  • “longer term highs & lows are more significant”
  • “it takes strong fundamentals to trigger longer term highs & lows”
  • “IMM Futures traders (COT speculators) follow 13, 26 and 52 week cycles”
  • “longer term channels should avoid more chop”

we stuck with a 90 day Donchian Channel as a trigger (Rule 1) with a 2×100 day ATR Trailing stop (Rule 2). But we wanted to make the strategy more robust, and take breakouts that were based on constructive price-action and not volatility-based breakouts. We did this by incorporating a simple trend-filter.

Only take long (short) breakouts when the 50 Day Simple Moving Average is above (below) the   100 Day simple Moving Average.  (Rule 3)

By doing so, we are effectively requiring that price, on average, has been closing higher (or lower) over a good length of time (being constructive).  By chosing a trend filter that is longer than the Donchian Channel, it becomes the dominant concept and the Donchian is simply the trigger. By chosing a trend filter that is shorter than the Donchian channel, the dominant concept remains the high/low trigger.

We’re going to have a time-stop based on the concept of instant gratification. The best trades are tendentially the ones that move in the desired direction quickly. So we want to scratch any positions that are not in profit after 10 days, even if they have not hit the stop loss. (Rule 4)

Here is what the final strategy looks like on a chart:

donchian-final

EurUsd Donchian Strategy  as programmed by Craig Consulting

Full Expert Advisor for MT4 is available for FXRenew Members (Free Trial Available).

What are the statistics of this 4-Rule strategy, across currency pairs and commodities that most retail brokers offer easy access to?

results

Results obtained with our robust, non-optimized 4-Rule strategy.

The caveats to these results are that the position size is fixed fractional, and not volatility based; there is limited diversification; there is no rebalancing at the end of the month (which instead is mainstream within professional money managers). 

Over To You

In this article, we have demonstrated that markets possess cycles of mean-reversion and continuation, but they do not just depend on volatility cycles.  Fortunately, continuation strategies like the good old Donchian Channel strategy still obtain satisfactory long term results across asset classes.

We have created a simple, non-optimized, 4-Rule Donchian Breakout strategy in MT4 (available for free to FXRenew Subscribers) and we have demonstrated that the concept of trend following is much more important than the entry rules themselves.

So when creating a strategy, the market type is the overriding variable. Position size and risk limits come next. And then, the key is to diversify across various asset classes and not stick to one thing (which generally leads to optimization and ultimately a loss of robustness). Focusing only on the entry will ultimately lead to a slow bleed of your account.

About the Author

Justin Paolini is a Forex trader and member of the team at  www.fxrenew.com, a provider of Forex signals from ex-bank and hedge fund traders (get a free trial), or get FREE access to the Advanced Forex Course for Smart Traders. If you like his writing you can subscribe to the newsletter for free.