Introducing the FX Trend Snagger model

This article describes the FX Trend Snagger trading model, a model designed to capture profits from trending behaviour in the currency markets. The model traded its first live position (short AUDUSD) on 2 October, 2012.

Unlike many trend-following models, the “snagger” logic of the model is designed to capture profits when trends only prove to last for a relatively modest period of time.

The weakness of many trend-following systems is that they only do well in cases of extremely persistent and long-lasting trending behaviour and often only exit a position once the price has gone clearly (and often extensively) against the established trend, leaving huge potential profits on the table. 

The intent of the Trend Snagger model is to provide a more consistent return profile and fewer feast-or-famine periods over time and to “snag” profits once the profit draw-up has crossed key thresholds and then momentum fades, rather than waiting for a countertrend move to reduce a position. Thus, this model will tend to underperform classic trend-following models when there is an exceptional trend in the market, but will outperform such models in cases of more modest trending behaviour.

A rules-based model
The model is rules-based with some minor discretionary leeway on the actual price levels where initial and maintenance stops are in place, as well as when profit is to be taken once a trend has become well-established. But regardless of the leeway in some rules that depend on market conditions, the basic risk (assuming orderly markets) for any single position is always the same from the time of trade entry, regardless of the distance to the stop as position sizes are scaled according to the potential loss on the initial stop as described below.

Instruments traded

At present, the model surveys the following instruments for signals:

- EURUSD, GBPUSD, AUDUSD, EURJPY, GBPJPY, AUDNZD

Trade Entry
The FX Trend Snagger model initiates positions using the following basic trigger:

1) EMA cross-over. Two exponential moving averages (short and long EMA’s) are used to define the trend, which is bullish when the short EMA is above the longer EMA and bearish when the short EMA is below the long EMA. A position is entered on the day (as measured at the closing price) the short EMA crosses above or below the long EMA.

2) Exceptions. If the last two signals (one in each direction) for an instrument resulted in stop-outs and there was no significant draw-up in profits during those trades, a third or subsequent signal is only taken once the market has moved out of the range established by the market over the course of the last two (or more if the market continues to churn in a range) previous trades – i.e., once the market closes outside or breaks out of the previous range. (see chart below).

Examples of trade entry signals not taken – the chart above shows EURJPY during the summer of 2011. Signals 1 and 2 would both have been taken and resulted in stop outs with no appreciable draw-up in profit during the life of the signals. Signals 3 and 4 should not be taken as the market remains range-bound and never closes outside of the range established by the failed signals 1 and 2. Signal 5 would be taken as the market closes outside of the previous range on the day of the signal.

Trade Exit
Once a trade has been entered, the model defines an initial stop that is a percentage of recent volatility. That percentage varies depending on the instrument pair traded and its historic behaviour. Eventually, a maintenance stop takes over once the draw-up of the position has moved quite well into profitable territory.

Stops
• Initial stop: (from approximately 1 to 1.4 Average True Range (ATR), depending on the instrument, usually about 1.0 ATR for EURUSD and EURJPY and higher for the other instruments the model trades. There is some room for discretion on the stop level if there has been a particularly large move in favour of the new trend before the trade signal actually occurred, or if the ATR is exceptionally small relative to the historical volatility of the pair traded. Regardless, the position size will be adjusted such that the distance to the initial stop will always result in the same absolute percentage loss (not counting slippage). This is described more thoroughly below, but the idea is to have a constant initial loss risk of x% for any new position regardless of the distance in pips or percentage of the underlying instrument to the stop.

• Maintenance stop: Once the trend becomes established – moves into profit by more than a certain ATR factor - the stop is adjusted to a new “maintenance stop” that is moved higher from the initial stop, but tends to widen a bit as the trade goes more significantly into profit. The jump from the initial stop to the maintenance stop is fairly large, but doesn’t necessarily take the stop into profitable territory. The maintenance stop is defined algorithmically and is generally not altered when a full position (the initial two-unit position taken) is still active. Once profit has been taken on the first unit (half) of the position, however, the maintenance stop is abandoned in favour of a slightly more discretionary stop that is often even wider than the standard maintenance stop to allow more room for volatility. Nonetheless, the stop on the second unit will always be placed to ensure that the second unit is still stopped out at a profit in case the trend doesn’t continue to develop favourably for the position.

• “Stops” due to trend change. On some occasions, the trend will change before the initial or maintenance stop levels have triggered. In such instances, a position would obviously be taken off and even reversed in favour of the new trend as defined in the Trade Entry section above.

Take profit exits
The system includes a fairly simple profit-taking logic that is clearly defined for the first half, or “unit” of the position. The rules for taking profit on the second unit are slightly more flexible and varied depending on how the market develops. The key for profit taking is the overall development of the model’s Trend Strength indicator, which is merely a statistical measure of where the price is versus the two moving averages and how stretched the strength of this measure is versus the recent historical record.

• Taking profit on first unit: There are two requirements for taking profit on the first half of a position. First, the draw-up in overall profitability must exceed a take profit threshold for the position that is a multiple of the instrument’s volatility. Secondly the trend strength indicator must be in its “second wave” of strength after reaching a minimum threshold of strength on a first wave (the first wave is defined as a run-up in the daily trend strength indicator – the wave is complete once a peak is reached and it begins receding again) and profit is only taken once there is a “pullback” day that represents a loss of momentum. A special stop on half of the position (the first unit) is employed in cases where the market has run well beyond the take profit threshold to ensure that a good profit is taken before the market reverses too much of the run-up in profits. See the example below for more.

• Taking profit on second unit: The requirements for taking profit on the second unit are a bit looser, though the logic is essentially the same. The profit threshold is basically twice that of the profit threshold for the first unit. A similar attempt is made on the second unit to wait for two waves of trend strength and then a day of fading momentum before taking profit. If a very large profit was taken on the first unit, the hope is generally to seek a larger profit on the second unit as well, though this is often naturally the result anyway if the trend continues and the model waits for the requisite trend strength patterns before taking profit.

• Exceptions. Exceptions to the above may be employed in cases where profit runs up in parabolic fashion very rapidly, well beyond the first and second thresholds, in which case profit may be taken on one or both positions as long as the units are well beyond their basic take profit thresholds.

Chart – Example of how the FX Trend Snagger is meant to work: Note that the model was stopped out on the first two signals (first up and second down arrows on the left) on the chart above and therefore doesn’t get long on the second long signal until the first green circle, when it breaks above the previous minor high. At circle 2, the momentum has taken the drawup to beyond the first nominal profit target, but the model doesn’t take profit until the next day when the price falls a bit, because momentum drops off on that day. The same goes for the second profit-taking trade at the circle marked 3. Note that in both cases, because the price has crossed above the nominal profit target a close stop is established in case the market retraces too far, in which case that unit would have been stopped out in profit at the red line indicated. This method is employed in case the trend continues to progress with no loss of momentum far beyond the nominal profit-taking targets – part of the overall “snagger” principle. At circle 2, for example, had the market continued to run sharply higher for two more days, the model would have stayed long for those two days, provided the trend strength indicator continued to run higher as well, which it generally does on days in which the trend continues strongly.

Re-entry on existing trend

Re-entering after a stop-out. On occasion, a stop-out occurs due to a choppy market and the instrument subsequently shows signs of trending again in the same direction, rather than reversing to a new trend in the opposite direction. For these situations, the model allows for re-entry on the original trend. The rules for re-entry require some room for interpretation, but the idea is to re-enter either if 1) the market merely spiked briefly to hit the stop and then immediately reverses back far away from the stop or 2) once the market closes at or near a new extreme in the favour of the trend. If, for example, a long position is stopped out, but then the market turns around again in subsequent days and heads back higher, the model will go long again on new close back at a new 6 to 10 day high close, depending on how range-bound the market has been. See the chart below for an illustration of the re-entry principles.

Example of re-entry after a stop out. EURUSD: The downward pointing red arrow shows the day on which the model detected a new down-trend and sell signal. Unfortunately for the model, two days later, the position would have been stopped out if using a 1 ATR stop (indicated at the red line within the circle), which was about 200 points at the time. Subsequently, the model allows re-entry in either of two ways: on the close of the day after which the position would have been stopped out (the day after the spike would have stopped the position out, as the stop-out clearly occurred on a temporary spike) or alternatively, the model can wait until a new low close as shown with the green circle two days late.r
Other re-entries – re-entry on major trends.

On occasion, if there is an especially long-standing trend and there is no position remaining in the model because profit has been taken on both units, a position of half size may be re-entered on a countertrend move, provided that the entry level is significantly better than where the profit was taken on the second unit of the original trending move.
A half or full position may be “re-entered” in cases where an EMA cross-over to the opposite trend very nearly occurred but then the market begins trending again in the direction of the last signal (for which profits have presumably already been taken and there is no position). Risk parameters are the same for these position types.

Trade Sizing
The position size for a new trade is called “two units” because profit taking always takes place in two phases, while trade entry always occurs at full size.

For all trades, the trade size for any new position is scaled according to the recent volatility of the instrument and therefore according to the distance to the initial stop level (which is also volatility-based). This is done so that regardless of the volatility of the current market or the underlying instrument traded, the risk for any new position is a constant. The different instruments the model trades have different initial stop tolerances that are based on the historical volatility of the those instruments – so EURUSD and EURJPY have approximately 1 ATR stops while other instruments stops are somewhat wider. This also means that the position sizes are relatively smaller for those other non-EUR pairs relative to their daily volatility. Note that slippage and/or weekend gaps can always mean a slight deviation from the aim of keeping a constant risk if there is a fast market or a gap over the weekend.

Portfolio Considerations
In addition, portfolio considerations demand that some position sizes are reduced (halved) or even ignored in the event that the model generated multiple signals for highly correlated instruments, particularly if those signals all arrive on the same day.
Important notes/exceptions.

Exceptions to any of the above may be employed in trading the model in cases of extreme volatility triggered by things like natural disasters (such as the Japanese earthquake/tsunami of March 2011) or other sudden events that trigger wild parabolic moves in a market. Any such exceptions are exclusively in the interest of avoiding putting on added risk or taking off existing market risk if the market has moved in such instances in favour of the positions (Stops are in the market, so if extreme events occur while a position is on, the market stops will be in effect, with the only added risk being one of slippage). In such cases, a few days will be needed to allow things to settle before new trading signals are accepted as actionable.