Introducing the Saxo Bank Turtle Model
The Saxo Bank Turtle Model made its first trade on September 13, 2012, with an initial investment of USD 65,000. The model trades in 15 instruments:
Equities: S&P 500
Softs: Coffee, Sugar, Cocoa
Metals: Gold, Silver, Copper
Energy: Crude Oil, Heating Oil, Natural Gas
Currencies: USDCHF, EURUSD, GBPUSD, USDJPY, USDCAD
Our Turtle Model is based on a trend following trading system developed in the 1970’s that gives signals for trade entries, stops and exits, but also rules for risk management. The idea is to catch breakouts in major liquid financial instruments and ride the momentum until the trend slows down.
The original model traded futures, but our model uses CFDs and spot instead. Furthermore, the original Turtle Model traded in 21 futures instruments, but due to obsolescence of some of the instruments and trading constraints on the Saxo Trader, we have narrowed our model universe down to 15 instruments.
Every day the model calculates a 55-day Donchian Channel. If the last day’s close is higher or lower than the upper or lower band of the Donchian Channel, a long or short position is initiated.
The model calculates every day a variable called N, which is essentially the 20-days exponentially moving average of the last trading day’s average true range (ATR). Basically it is factor that describes the current market volatility.
To standardise positions the current volatility, defined by N, is converted to dollar volatility by the formula below. This adjustment is essentially only relevant for CFDs as they are based on the underlying futures contract which dollars per point is typically more than one as is the case with spot instruments such S&P 500 and Gold Spot (XAU).
Based on dollar volatility, the model then calculates something called Units, which is essentially the final step to volatility adjusting the positions. Units were sized so that 1 N represented one percent of the account equity.
As the portfolio can be quite volatile in its original setup if many positions are added, we have reduced the equity allocation per unit to only 0.25% of the account value.
Adding units following entry
The model enters a single Unit at the breakout and adds to this position at 0.5 intervals of N until the maximum limit of four Units in the same instrument is reached.
Example: Natural gas, N = 0.25, 55 day breakout = 3.000
If the Unit Size in the natural gas example is 1,000 Mmbtu, the models adds 1,000 Mmbtu at every price point for the next Unit (1,000 Mmbtu is a trade lot worth USD 2,815 as of 5th of September, 2012).
The model always has stop orders on open positions. The Turtle Model places stops based on position risk. No trade could incur more than two percent risk (remember that one N represent one percent position risk).
Stops are thus placed 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 were added, the stops for earlier units were raised by 0.5 N. All units are stopped out at the same price.
Example (continuing with natural gas):
This is essentially taking profit. The signal is based on 20-day Donchian Channel. If our long position in natural gas (see previous examples) has moved up to 3.500 with the upper bound of the 20-day Donchian Channel @3.390 then our exit signal is @3.390 and at the same time above the stop @2.875. All units are sold when the exit price is reached.
Some positions do not move “deep” enough into profit territory for the upper bound of the Donchian Channel to move above our stop. If our long position moves down and reaches our stop the position goes from profit into loss territory and gets stopped out. So for exits to be used, the positions have to trend up or down for a considerable period.
To avoid excessive exposure to individual markets, the Turtle Model also has positioning limits depending on how tightly correlated markets are with one another (see table below)
It is not specifically defined what 'closely' and 'loosely' means in this context, but we will use the rolling 60-month correlation matrix (see below) as our guideline for determining the degree of correlation. We have decided that instruments with correlation below 0.5 are loosely correlated and above are closely correlated.
Source: Bloomberg L.P., Saxo Bank Strategy & Research (correlations as of 5th of September, 2012)
In all CFD positions, the model has to roll its positions before the underlying futures contract expires. This is done before the volume and open interest have declined too much.