Systematic Breakouts – Turtle Style By Justin Paolini

“I’m going to grow traders like they grow turtles” – Richard Dennis, Market Wizard

Almost everyone has heard of the Turtle Traders. Whether through Curtis Faith’s “The Way of the Turtle” or from online forums, the Turtle Trading Model has been written about extensively.

So in today’s blog post we will explore both the mecchanical application of the system itself but we will focus on 2 components of the system that make it different from other channel breakouts:

• the P/L filter;
• pyramiding into position.

The Turtles

This article in reality builds on 2 previous articles that you can read here and here.

In this article we are not only exploring the feasibility of the Turtle Trader’s method, applied in a systematic manner, but we are also going to explore the 2 peculiarities of the model:

• the P/L filter (don’t trade the next 20Day breakout unless the previous was a loss);
• pyramiding into position (Adding positions every 1/2 ATR up or down the ladder).

What has been left out from traditional discussions on the Turtles is that the Turtle Trading System was never meant to be traded on a purely mechanical basis. Richard Dennis made a bet with William Eckhardt that traders could be trained (Eckhardt felt that successful traders were “born”, not “nurtured”).

Therefore, he did not present the mechanical trading rules alone: there was substantial discretionary criteria presented as well. But before exploring the discretionary criteria, let’s see how the system acts in today’s markets. Once again it’s tip of the hat to our resident programmer Tony for going through all the trouble of coding this.

Our tests were performed on WTI (from 2011), US30 (from 2008), AUDUSD, EURGBP, EURUSD, GBPUSD, NZDUSD, USDCAD & USDJPY(from 2000). The number of scale-in entries was varied from 1 – to 4, each set 1/2 ATR apart. The Turtles used 4 entries, spaced 1/2 ATR apart.

It is truely surprizing to see just how different the end results can be, by compounding into a move. Here are the numbers:

With this information now at hand, we can go back and evaluate whether pyramiding functions in the same way with simple breakouts and with any breakout length.

P/L Rule Evaluation

This was an even more interesting test because the Turtles were taught to trade a 20 day breakout only if the previous 20 day breakout was a loss. I’ve tried to rationalize this and so far have been unsuccessful. A couple of ideas that pop in mind are:

• attempting to avoid whipsaws;
• an empirical observation that successful breakouts (i.e. longer trends) rarely happen in succession;
• an empirical observation that after successful trends, the market tends to consolidate before making the next move.

In the chart below we have the original system which uses the P/L filer (Prior Break). System 1 is the 20 day breakout without the P/L rule. System 2 is the 55 day breakout without the P/L rule. 50/50 blend is an equal weight of system 1 and system 2.

The P/L filter seems to be some kind of “trend quality” filter. The number of trades is reduced, if compared to the blended system. The win ratio is not statistically different however, in either model. So where is the catch? For now, until we can understand the dynamic better, it does appear to be simple empirical research done by Richard Dennis.

Qualitative Filters

Richard Dennis also suggested qualitative screening for the best trade ideas.

• One of the most important filters for Dennis was a consolidation on a chart. The volatility contraction, alongside the evident range, makes continuation more logical after a breakout.

• Hold consolidation during adverse news. The Turtles did pay attention to fundamental information about the markets they were trading.  So, if they were monitoring GBP, and a news suggested a worse Employment report, but the market failed to sell-off, the Turtles would take that as a positive signal for the market.
• Momentum Rules like rallies on bad news (for a long position) or selling on good news (for a short position), recent false breakout in the other direction (P/L Rule), visual gain in momentum.

There were more rules but they fell under the “minor” category and as such we will dismiss them here.

Over to You

Simple ideas continue to work in the markets, to this day. When everyone is screaming that trend-following is dead, it’s probably the right time to try out a trend-following model!

Even with limited diversification, we have found the model to have a surprizing SQN even in today’s markets.

The interesting attributes of the Turtles’ model are the P/L filter and the pyramid structure. They both seem to add value in a way that we still need to understand better. For now, the main takeaway is that there is no need to reinvent the wheel in order to be a profitable trader. Good inputs do seem to be evergreen or timeless!