- Equities
- Interest rates
- Agricultural commodities
- Non-agricultural commodities
- Currencies
- The look-back period: short-, medium- or long-term trend-following
- The sensitivity of the trend signal: how frequently we look at prices (e.g. daily, monthly)
- Diversification across parameters: It would be better to combine short-, medium- and long-term signals. Alas that tends to increase capital requirements as often several futures contracts would be held per asset.
- Non-binary entries and exits by scaling in and out of positions would be preferable in practice resulting in a smoother equity curve. Using different time frames could achieve that.
- Taking correlations into account for diversification: the benefit of adding an uncorrelated asset far outweighs adding another closely correlated one. We can use broad categories (e.g. for equities only US, Developed and Emerging markets) and cap closely correlated assets to achieve good diversification with a lower number of positions.
- The basic strategy is common knowledge and a lot of the underlying market inefficiencies are likely to be arbitraged away by increasing participation when many trend-followers enter and exit at the same time. A useful idea would be to implement a counter-trend element that uses that insight by entering near common trend-following exit points.
- Investing in smoothly trending markets is best in practice, but that is hard to quantify and backtest as well as it is unpredictable when a smooth trend phase is likely to begin.
- Carry (the cost of holding a position) could add useful information on expected returns for different assets and add an additional dimension especially for commodities and currencies.**
- 9-day moving average: 9MA
- 60-day moving average: 60MA
- 275-day moving average: 275MA
- new long-term high or low
- Long-term trend component: if 60MA is above 275MA enter only long positions; if 60MA is below 275MA enter only short positions.
- Discretionary trend quality: I visually determine if an asset currently shows smooth, clear up- or downtrends or follows a choppy path with lots of gaps to whittle down the number of investable assets: I assume here that trend quality exhibits persistence which will result in more profitable trades, because less initial stops are triggered by random movements, but, as it isn´t easily testable, who knows? I base my assumption on the observable phenomenon of volatility clustering.
- Carry: Holding an asset should cost less than 1% per quarter, positive carry positions are preferred. Especially for commodities this is an essential screen, futures for most other assets show very small cost of carry that we can ignore.
- Counter-trend, medium term component: after a counter-trend that causes the 9MA to cross below the 60MA enter an initial position in the direction of the long-term trend when the 9MA re-crosses above the 60MA. This is a high probability entry that buys outsized dips below recent highs (or the reverse for short positions), but only enters in the aligned direction of the long-term and medium-term trends.
- Long-term trend component: enter a second position at the breakout to a new high or low. The setup of the strategy will usually cause a new high/low to be several weeks from the last significant high/low. The initial position must have moved well into profitable territory at the second entry point or we don´t take it.
- Short-term trend component: exit one position at a daily close below (for longs) or above (for shorts) the 9MA. This exit determines the actual capital at risk for each position and it will closely follow a price move in the right direction.
- Medium-term trend component (only when two positions were entered): exit remaining position when the 9MA crosses below (for longs) or above (for shorts) the 60MA.
- The exit point on the other side of the 9MA initially determines my average potential loss.
- I volatility weight each position by calculating: 0,33% of capital (this percentage is determined by my risk preference and portfolio structure) divided by the current 20-day ATR multiplied by the futures contract multiplier which gives me the number of futures contracts to buy or sell:
(0,33% NAV) : ((1 ATR) x contract multiplier) = number of contracts.
I then cross check, if my 9MA exit at the time of entry is approximately 1 ATR away from my entry point, as this will determine my average loss in case the position goes against me right away – 0,33% of capital are at risk for each position using the sizing formula. If the distance between entry and exit is further than 1,5 ATR, I don´t take the entry, because the initial risk is too high. - When I get a second breakout entry signal I follow the same procedure for the second position while putting a mental breakeven stop loss below my first position, so I have never more than an average 0,33% of capital at risk per asset.
- At the same time I have an eye on the capital at risk by all open positions simultaneously. This should be around 5% at a maximum. When many assets show entry signals simultaneously, I implement a correlation cap – e.g. after three positions I stop adding any more equity markets as they are very closely correlated.
- Each position exhibits the same average day to day volatility (an average daily change in value of 0,33% NAV) – portfolio risk is diversified across uncorrelated assets.
- Whether a position will be a loser is determined quickly because of the tight stop loss. This is likely the weakest point of the strategy: if too many positions are stopped out by random noise, it will not be profitable.
- Profits will be taken quickly initially and only the positions that are profitable and have their trend confirmed by a breakout to a new high or low will be converted into medium-term holdings through the second entry and exit points. Holding on to these profitable positions long enough is the most challenging part in running the system for me, but it is where the real profits are generated.
- Because of the tight initial exit and the strict entry criteria, that weed out a lot of assets, we will likely hold only a small number of open positions at any one time (4 to 12) each at a larger size than a traditional CTA would scale it (they would use about 0,1% to 0,2% NAV in the above formula*), which allows the strategy to be implemented with less capital then is usually required.
- The volatility of profitable positions may be quite high as position sizes can be comparatively large due to the tight exit and pyramiding mechanism. A high upside volatility is desirable, but considerable book profits can also revert quickly and be stopped out.
- The size of the positions makes the use of leveraged instruments necessary – only a handful of positions will use up all available capital, if used unlevered. Leverage levels vary according to volatility (short term interest rates will require most leverage), but can easily reach and exceed 10 times overall.
Managed Futures in a portfolio context
It is not easy to judge how much capital to allocate to the strategy within a portfolio as in practice futures contracts only use a part of your margin allocation (usually between 10% and 20%) and no capital. The actual risk allocation in the portfolio is determined by the combined position volatility which is determined by the position sizing formula.
I arrived at the position sizing algorithm above by combining backtest data, comparing margin requirements of futures positions with the margin used by my short options strategy and the exposure to the underlying assets (where managed futures inherently use a higher degree of leverage). I now watch the daily fluctuations within my portfolio closely as well as monitoring capital at risk in different portfolio parts independently to fine-tune the allocation. Updates are listed below.
My default allocation in the current positive market environment is equal weighting exposure to the three basic strategies in my portfolio:
global asset allocation – short volatility – trend-following managed futures
The concept of adaptive portfolio allocation will change the weight with changing market regimes, quickly reducing exposure to short volatility strategies in a beginning bear market, more slowly reducing exposure to and reallocating within the global asset allocation, while raising cash and increasing the managed futures allocation.
Adjustments
January 26th 2018: After running the strategy for two months with good results, I reduced the daily average volatility per position from 0,5% of capital to 0,33% of capital. I found myself holding around 10 positions on average and, while usually the assets move uncorrelated to each other, occasionally everything co-moves. This has caused daily fluctuations in the portfolio value of almost 5%, which I find too high.
I also reduced the ATR look-back from 60 days to 20 days to react more quickly to changes in volatility. Starting in February 2018 volatility shot up across many assets and open / new position sizes should have been adjusted faster.
August 2018: February´s regime change introduced increased whipsaw losses to the strategy and I looked at smoothing and enhancing returns by combining the trend strategy with some mean-reversion ideas based on carry. These can be found in a new post.
14 comments