Double Your Treasure : Two-filter trading strategies use a simple technical screen to point traders in the right direction

Less Equals More

When constructing a model for analysing and trading a complex entity such as the financial markets, the fewer the parameters used, the greater the model’s ability to generalise for unseen cases. This is known as statistical robustness and represents the ability of a model or a system with a fixed structure to hold up in a constantly changing environment. We simply don't know how the future will unfold. To make the best of this bad situation, the correct approach is to build strategies that attempt to capture the features of the bulk of possible future scenarios, rather than all of them. Strategies that rely upon many layers of rules and exceptions yield less-than-optimal results in a world where the underlying assumptions are constantly changing. The strategies presented herein rely on few parameters, making them not only easy to memorise and implement in the heat of the battle, but also robust.

Not Only Direction

Traders often refer to a “good trade” as one that provides rapid confirmation of their directional biases. Thus, from a trader's perspective, simply being on the right side of a market is not the sum total of a winning trade. Even if price direction has been correctly predicted, your position still needs to appreciate enough to cover your costs (i.e., spreads, commissions and losing trades), and at a rate above the risk-adjusted return of other opportunities. All else being equal, the less time it takes to cover trading costs and exit the trade with a profit, the better. Flat price traders therefore not only want to select market situations that support their directional biases, but that also lead to an increase in short-term volatility. The two-filter strategy attempts to achieve both of these goals.

Entry Filter

Significant price moves – i.e., moves that qualify as more than just “noise” – are usually interrupted by periods of rest before continuing in the direction of the trend. Sometimes these rest periods are deep and protracted, at other times, brief. Traditional chartists use the phrase “flag” or “pennant” to refer to the short-term variety of rest periods that occur in the context of strong moves. Let’s take a look at some examples of these classic patterns.

Figure 1 is a chart of the June 1985 gold contract. After declining over $10 per ounce from its April high, the market paused for four days before resuming its downward motion. The eventual move below the lower bounds of the pattern was rapid and substantial.

Figure 2 is a chart of April 1996 crude oil futures. Following a sustained rise into late February, supply is finally induced, causing the market to halt its advance. But demand is still sufficient to maintain crude oil prices between $19 and$20 per barrel for almost two weeks. Once supply has finally been absorbed, crude oil moves sharply higher into March. These are examples of classic “flag” patterns.

However, there is often considerable variation in the length of time it takes each pattern to form, in the angle of the pattern and in other aspects. In fact, we can probably find an infinite variety of patterns fitting the general description of “brief pause.”

To arrive at an objective substitute, then, what main components can we distil from the general idea? Two that come immediately to mind are:

1. The concept of prices trading briefly in the opposite direction of the main price trend

2. The concept of a rapid increase in range,

i. e., short-term volatility.

To capture the first component, we merely require a single period (i.e., one hour, one day, one week, etc.) where the closing price is below the opening price for a bullish setup, or where the closing price is above the opening price for a bearish setup.

The second component, a rise in short-term volatility, is captured by requiring the market to trigger an entry within two periods (i.e., within two hours, two days, etc.) following the initial setup period. Entries are triggered as follows: following a bullish setup, the market must open below the high of the setup period. For bearish setups, the market must open above the low of the setup period. Trades are entered only if price trades up and through the high of the bullish setup for long entries, or down and through the low of the bearish setup for shorts, and only within the two periods immediately following the setup.

If the market gaps open above the bullish entry point (or gaps down, below the bearish entry point), the setup is discarded. If for any reason no trade is triggered within the two periods allotted following the setup, the setup is abandoned.

By requiring that prices fully retrace the setup period's open-to-close direction within two periods after the setup occurs, we are asking the market to prove - in short order -that all of the previous period's buyers (or sellers) were mistaken. The speed and extent to which prices rise or fall after a trade has been entered help answer the question of whether there is power behind the move, and whether buyers (or sellers) are acting with conviction. Either short-term volatility expands - in the form of prices surging past the entry point - or it does not. The answer should arrive quickly and decisively, or not at all. A schematic of this entry technique can be seen in Figure 3.

The protective stop for long entries is the lowest low of the entire pattern, or the highest high of the pattern for short trades. Stops are moved to break even once open profit equals the initial risk (simply defined as the difference between entry and stop price). Once the stop has been moved to break even, all or part of the position can be exited at multiples of the initial risk taken. In my own testing, I find that among the trades that survive the break-even stop, greater than half of these trades go on to exceed 200% of the initial risk amount.

Trend Filters – The Quick and the Slow

The next step is to combine the entry pattern with a trend filter. Long entries are acted on in “up” trends and short entries are acted on in “down” trends. By definition, trend filters are lagging indicators. Trend filters help answer the question “which way is the market moving,” but do not solve the mystery of when one trend is over, and another has begun. There is always a trade-off between “too fast” and “too slow.” Shorter-term filters produce earlier indications but include many false alarms and aborted trends. Longer-term filters produce later indications but fewer false alarms and more sustained trends. My approach to this, after studying the problem for a number of years, has been to settle on two very simple filters, one quick and one slow.

The faster acting trend filter defines any market that closes above its 10-period simple average as an uptrend, and any market that closes below its 10-period simple average as a downtrend. The slower trend filter designates any market making a recent 30-period high or 30-period low as being in either an uptrend or a downtrend, respectively. Recent simply means within the last five periods.

Let's review some examples. At point F in Figure 4, the close of the July' 05 CBOT wheat contract is below the open. The shorter-term trend's filter is “up,” since the price is above its 10-period average. The high of the setup bar (coloured blue) becomes the entry point if taken out within two days. In this example, wheat immediately takes out the previous high. The result of the trade was positive – assuming you moved your stop to break even and exited the position after prices reached a multiple of the initial risk.

Another bullish setup occurs at point G, but no trade was taken, as the market did not confirm an entry within two periods after the setup. Setup H yielded a successful long entry, taken in the context of a recent, 30-period high. A losing long trade would have been entered and stopped out for a loss at the low of the setup bar at point X. There was no trade triggered following setup Y. Lastly, a bullish setup occurred at point Z. The market confirmed an entry by trading up and through the high of the setup bar within two periods following the setup.