This report is produced by Huobi Research; please cite “Huobi Quant Academy” for reference.
The Turtle Trading System is a classic, comprehensive and long-lasting trading system, one that has influenced countless of investors 40 years after its invention. The system includes many classic concepts such as trend spotting, position sizing, risk adjustments and stops/exists, and has helped investors to become better traders and win them sizable investment returns in their careers. Cryptocurrency is the next emerging market after financial derivatives such as options and futures; as the development of Blockchain technology continues, this asset class will certainly draw attention from more investors in days to come. This article applies the traditional Turtle Trading system to Cryptocurrency trading. By using quantitative trading strategy to backtest the performance of turtle trading system in BTC market, this report thoroughly evaluates the effectiveness of this classic in a brand new field.
1. Background of the Turtle Trading System
The Turtle Trading System originates from a famous dispute in 1983. Famous commodities speculator Richard Dennis had an ongoing dispute with his long-time friend Bill Eckhardt about whether great traders were a god-given talent; Richard believed that he could teach people to become great traders, while Bill thought that genetics and aptitude were the determining factors. In order to settle the dispute, Richard suggested that they recruit and train some traders before letting them trade on actual accounts. Since Richard was a famous trader at the time, he received submissions from over 1000 applicants. In the end, Richard recruited 13 volunteers to proceed with his experiment, including 10 novice traders and 3 friends that Richard already knew. The group of 13 people were invited to Chicago and trained for two weeks at the end of December 1983, and began trading with small accounts at the beginning of January 1984. After the training period ends, Richard funded most of them with $500,000 to $2,000,000 accounts at the start of February, and the students were referred to as the “turtles”. It later become one of the most famous experiments in trading history because after four years, these “turtles” achieved an average annual compound ROI of 80% with their given accounts. As such, Richard successfully proved that, with sufficient training, ordinary people with no prior trading experience could also become excellent traders.
Several years later, some of the successful “turtles” decided to publish the turtle trading rules, which subsequently benefited numerous of investors around the world. The Turtle Trading System is essentially a comprehensive trading system that covered five most important aspects of investment:
- Markets — What to buy or sell;
- Position Sizing — How much to buy or sell;
- Entries — When to buy or sell;
- Exits: — When to get out of a winning position;
- Stops — When to get out of a losing position.
In short, the Turtle Trading System bears three distinguishable features:
- It is a comprehensive trading system that covers every aspect of trading;
- Traders simply follow the instructions without making any subjective judgments.
- Due to its proved profitability, traders rarely doubts the authenticity of the rules, even in losing positions
To a certain extent, the success of Turtle Trading System could be attributed to its simple and automatic nature, since these two characteristics leave traders no room for subjective thinking. In the words of Curtis Faith, the author of Original Turtle Trading Rules, “Most successful traders use a mechanical trading system. This is no coincidence. A good mechanical trading system automates the entire process of trading and provides specific instructions for traders. The system consistently makes it easier for a trader to trade because of the set of rules that specifically outlines what should be done in every situation. As such, the mechanics of trading are not left up to the judgment of the trader.”
The success of the Turtle Trading System also reveals that, the confidence, consistency, and discipline offered by a thoroughly tested mechanical system are the “secret” to many traders’ success. Everyone with the confidence and the discipline to consistently apply the effective rules they were given can also become successful traders. As Richard Dennis, the father of the the Turtle Trading System once said, “I always say that you could publish my trading rules in the newspaper and no one would follow them. The key is consistency and discipline. Almost anybody can make up a list of rules that are 80% as good as what we taught our people. What they couldn’t do is give them the confidence to stick to those rules even when things are going bad.”
2. Comprehensive Guide to Turtle Trading System
Now, we will carefully analyze the Turtle Trading system based on its answers for the six questions listed below.
- Markets — What to buy or sell?
The Turtles only trade the most liquid markets. Higher liquidity leads to a favorable market depth, which can decrease trading risks and thus potential unwanted losses. Applying this rule to Cryptocurrency assets, this means that traders should focus on mainstream Cryptocurrency assets if they want to apply the Turtle Trading rules to their trading strategies.
- Position Sizing — How much to buy or sell?
The Turtle Trading system automatically adjusts the position size based on the Market Dollar Volatility, which measure the absolute volatility in dollar terms (when compared to positions in other markets)irrespective of the underlying volatility of the particular market.
Here we use N or ATR to represent the 20-day exponential moving average of the True Range.
Since this formula requires a previous day’s N value, you must start with a 20-day simple average of the True Range for the initial calculation.
The turtles build positions in pieces which we called Units. Units are sized based on volatility so that each N represents 1% of the account equity. Thus, a unit for a given market or commodity can be calculated as:
Since the Turtles use the Unit as the base measure for position size, and since those units are volatility risk adjusted, the Unit is a measure of both the risk of a position, and of the entire portfolio of positions. There are also risk management rules that limited the number of Units on four different levels:
Single Market (Maximum Units: 4);
Closely Correlated Markets (Maximum Units: 6);
Loosely Correlated Markets (Maximum Units: 10);
Single Direction — Long or Short (Maximum Units: 12).
In the Cryptocurrency market, traders applying the Turtle Trading rules can use correlation coefficient to determine the price correlation between Cryptocurrency assets.
- Tactics — How to buy or sell?
The Turtles strictly follow trade signals. As a mechanical trading system based on trend-observation, the Turtle Trading rule believes in “Buying Strength & Selling Weakness”. Moreover, in order to avoid unsystematic risk, the Turtles diversify investments and spread risk across different instruments. Additionally, they change instruments once in a while and when price changes, they also change investment instruments. The Turtles avoids subjective trades by following the systematic strategies, therefore they are told to strictly execute every single trading rule under all conditions, regardless of wins or losses.
- Entries—When to buy or sell?
The Turtles use a very simple entry system based on the Channel Breakout systems taught by Richard Donchain. The Turtles are given rules for two different but related systems we called System 1 and System 2, whereas System 1 is a shorter-term system based on 20-day breakout, while System 2 is a long term system based on a 55-day breakout. Investors are given full discretion to allocate as much as of their equity to either system at will.
A breakout is defined as the price exceeding the high or low of a particular number of days; Thus a 20-day breakout would be defined as exceeding the high or low of the preceding 20 days.
Turtles enter positions when the price exceeds by a single tick the high or low of the preceding 20 days. If the price exceeds the 20-day high, then the Turtles would buy one Unit to initiate a long position. If the price drops one tick below the low of the last 20-days, the Turtles would sell one Unit to initiate a short position.
System 1 breakout entry signals are ignored if the last breakout results in a winning trade. (A winning trade does not represent a real profit; it also includes a breakout ignored by 2 yet discovered to be profitable later.)
Turtles enter positions when the price exceeds by a single tick the high or low of the preceding 55 days. If the price exceeds the 55-day high, then the Turtles will buy one Unit to initiate a long position. If the price drops one tick below the low of the last 55 days, the Turtles will sell one Unit to initiate a short position.
Turtles enter single Unit long positions at the breakouts and add to those positions at ½ N intervals following their initial entry. This ½ N interval is based on the actual fill price of the previous order. So if an initial breakout order slips by ½ N, then the new order will be 1 full N past the breakout to account for the ½ N slippage, plus the normal ½ N unit add interval. The will continue right up to the maximum permitted number of units.
- Exits — When to get out of a winning position?
The Turtles follow a very simple exit strategy. The System 1 exit is a 10-day low for long positions and a 10-day high for short positions; All the Units in the position will be exited if the price goes against the position for a 10-day breakout. The System 2 exit is a 20-day low for long positions and a 20-day high for short positions; All the Units in the position will be exited if the price goes against the position for a 20-day breakout.
- Stops — When to get out of a losing position?
The Turtles place their stops based on position risk. No trade can incur more than 2% risk. Since 1 N of price movement represents 1% of Account Equity, the maximum stop that will allow 2% risk will be 2 N of price movement. Turtle stops are set at 2 N below the entry for long positions, and 2 N above the entry for short positions. In order to keep total position risk at a minimum, if additional units are added, the stops for earlier units are raised by ½ N. This generally means that all the stops for the entire position will be placed at 2 N from the most recently added unit. Since the Turtle’s stops are based on N, they adjust to the volatility of the markets. More volatile markets will have wider stops, but they will also have fewer contracts per Unit, This equalizes the risk across all entries and results in better diversification and a more robust risk management.
3. Backtesting Turtle Trading System
In order to evaluate the effectiveness of the Turtle Trading system as described above, we used Catalyst, an open-source project, as the instrument for backtesting, and used minute trading data on Bitfinex from June 30th 2017 to June 30th 2018, in which BTC price experienced a full Bull-to-Bear cycle. Since BTC volatility is significantly higher than commodity futures and stocks, and since BTC is traded 24/7, we believe that one year of trading data is sufficient for backtesting.
- Backtesting Model
We separate the backtesting model into four separate scenarios (by System 1/2 and by long/short positions). The four scenarios are described as follows:
- System 1 (Long + Short)
- System 1 (Long Only)
- System 2 (Long + Short)
- System 2 (Long Only)
The result of our backtesting model is shown in figures below. Each figure contains four sub-figures, in which the first figure represents daily portfolio value in USD, the second figure represents BTC price trend in the given time range, the third figure represents daily profit percentage of current strategy and the fourth figure represent daily cash value (negative value represents leveraging).
As a result, System 1 underperformed BTC price during the backtest period, while System 2 outperformed BTC price during the backtest period.
Based on the backtest results, it seems like simply following the Turtle Trading system does not yield significant profits, and that System 2 yields more profits than System 1. Under System 2, Turtles traded less frequently, thus having less stop-losses and an extremely low chance of entering a position during a downtrend.
What’s more, based on the backtesting result, we can see that shorting is probably not a great idea in Cryptocurrency market. Under the Turtle Trading rules, shorting does not benefit much and frequent stop-losses during a bull run leads to capital loss.
In essence, the Turtle Trading system is a trend-following strategy. As the backtest result shows, this system requires further optimization before it could be applied to Cryptocurrency trading. What’s more, although the original Turtle Trading system is intended for trading loosely correlated markets, the correlation in price between different Cryptocurrency assets are rather high — making it even harder to apply this system to the Cryptocurrency market.
In order to further optimize the performance of the Turtle Trading system in the Cryptocurrency market, We suggest the following adjustments:
- Use Moving Average (MA) channel as entries and exits
- Experiment with other time frequencies; such as 30 min, 2hr, 4hr and 6hr of trading data.
- Experiment with different stop placements in System 1, such as setting stop at 3N (or more) below the entry for long positions and above the entry for short positions. the baseline to 3 N or more.
- Experiment with different allocation of equity to System 1 and 2, for example: 40 % in system 1 and 60% in system 2.
We will continue to share the backtest result of proposed optimization in our future reports.
Based on the backtesting result of applying Turtle Trading system to BTC trading, we conclude that, despite its extraordinary performance in the traditional market, the Turtle Trading system does not yield a significant profit in the Cryptocurrency market. This is possibly due to the following reasons:
- High correlation in price of the Cryptocurrency market strongly affects risk distribution;
- Discrepancy in a Bull-to-Bear cycle between Cryptocurrency and other traditional asset so that the Donchian channel can no longer serve as an effective indicator in the Cryptocurrency market;
- The trading signals in the Turtle Trading System are well known by traders, thus reducing its effectiveness.
- Wes McKinney.Python for Data Analysis
Huobi Research of Blockchain Application (Huobi Research) was founded in April 2016 and started research and explorations in various aspects in blockchain area since March 2018. We cover blockchain technology research, industry analysis, application innovation and economic model explorations etc. We aim to establish a research platform and to offer theoretical foundations as well as judgements of trends in blockchain to the public, ultimately promoting the development of the entire industry.
Medium: Huobi Research
Facebook: Huobi Research
1. Huobi Research does not have any form of association with blockchain projects or other third-parties mentioned in this report that could jeopardize the objectivity, independence and fairness of this report.
2. All outside information, data referenced in this report is from compliant and legitimate sources that we deem as reliable, and Huobi Research have conducted the due diligence concerning its authenticity, accuracy and completeness, but such due diligence does not provide any guarantee.
3. This report is only for reference purposes. Conclusions and viewpoints in the report do not constitute any form of investment advice on crypto assets. Huobi Research is not responsible for any losses resulting from the use of this report, unless stipulated by law. Under no circumstances should the readers give up their own investment analysis and judgements.
4. This report only reflects the opinions from Huobi Research on the day it was finalized. Future market condition changes may lead to updates of such judgements.
5. The report is copyrighted by Huobi Research, please cite the source when quote, and get approval from us when large amount of contents is referenced. Under no circumstances is reference, abridgment and modification contrary to original intention permitted.