TSCI.pngAsymmetric Tradescapes

The tradescape technology can be used to determine the degree of asymmetry needed in a three-state signaling system.

If we assign +1 to a trading signal for a long position, -1 to a trading signal for a short position, and 0 for the state of being out of market, we can create a composite signaling system using two different EM lengths. We define asymmetry as the ratio of the EM length of the component generating the +1 signals to the EM length of the component generating the -1 signals.

For long trading, the +1 signal generates the entries and the -1 signal generates the exits. A favorable low value of asymmetry, such as 0.5, means that the entity prefers a quick to entry, slow to exit paradigm. This is typically observed with strong entities where traders are slow to abandon positions when drawdowns occur. A favorable high value of asymmetry, such as 2.0, means that the entity prefers a slow to enter, quick to exit paradigm. This is generally observed with weaker entities where traders rapidly abandon positions when a drawdown takes place.

The reverse is true for short trading where the -1 signal generates the entries and the +1 signal generates the exits.

Difficult to Signal Entities

Let's look at the long and short 10yr EOD tradescapes for CSCO:



The ten year robust trend for the CSCO underlying is a 0.45 CAGR. The RRt reward to pain is 0.02. The above tradescapes reflect the absence of any long term growth over the past ten years. The RRt scale is set to show only where more reward than pain occurs. We see that it is possible to realize profitable CSCO trading, even to see more reward than pain, but we also see that we must have an exceptional signaling system. The tradescapes clearly tell us that CSCO is a difficult entity to trade, long or short.

The basic tradescape uses a symmetric EM signaling algorithm. The time horizon used for the entries is the same as that used for the exits. This works well for entities that are fairly well balanced. For those entities the market treats as strong or weak, however, one wisely designs an asymmetric signaling system. This could be as simple as using two very different settings for entry and exit breakouts.

Tradescapes have an extension that we can use to determine the measure of asymmetry inherent in the market's treatment of an entity.

Asymmetric Tradescapes

The principle behind asymmetric tradescapes is that we can generate composite signals using two different lengths of the EM algorithm. The first component is used to generate the signal turns to the upside, and the second the signal turns to the downside. When the composite signal is in one of the states, it looks for a transition in the opposite state for its own change of state.


The full predefined set of asymmetric tradescapes are a matrix of 25 tradescapes with the following asymmetries:
1/10, 1/9, 1/8, 1/7, 1/6, 1/5, 1/4, 1/3, 4/10, 1/2, 2/3, 8/10, 1, 1.25, 1.5, 2, 2.5, 3, 4, 5, 6, 7, 8, 9, 10

The above is a long only set of asymmetric tradescapes for CSCO. The middle plot is the symmetric case, and it matches the long tradescape first shown. For long trading, CSCO is shown to be an entity the market treats as weak, a slow to enter, fast to exit entity. The optimal asymmetry is about 3.0. The good news is that CSCO can be traded successfully with an asymmetric system without having to resort to low-lag heroics.

We know that we want to build a signaling system where the ratio of information content used for the entries is three times greater than the exits. We want to be slow to enter, but once in a position, we have to be prepared to exit quickly.


If we look at the asymmetric tradescapes for CSCO short-side trading, we understandably see the opposite in the asymmetry. We need to be slow to enter any short position, but fast to close it out. There is usually far less order to trading downside trends. The downside movements tend to have much less order within them. Tradescapes will typically suggest that trading downside movements using any order-based system is either not going to work or it will be difficult. Here we see we have to be very focused in time horizon and have relatively low lag in order to trade short successfully.

Neutral Entities

Asymmetric tradescapes readily identify entities whose entry and exit behaviors are similar. It can be useful to know an asymmetric signaler is not needed.

This is a set of asymmetric tradescapes for QQQ, 10 years EOD, long only:


While there is certainly a different behavior with the two different sides of the asymmetry, one would be hard pressed to not choose the symmetric tradescape (the one in the center of the 25 tradescapes) as the optimum.

Forgiving Entities

Asymmetric tradescapes can also be used to find an edge with entities that are more easily signaled. This is the 10 yr EOD long tradescape for AAPL:


While some of the split patterns might raise concerns, there are clearly certain instances of fast to enter, slow to exit asymmetries that accommodate a weak or higher-lag signaling algorithm. Keep in mind that it is possible to retain close to full accuracy (with respect to trading the order in the price movements) with a 1.0-1.2 lag fraction signaler, where it is very difficult to achieve such accuracy at lower lag fractions. Any approach that opens the way to use a 1.1 lag fraction signaling algorithm also opens the potential for a very high accuracy signal as well.

Asymmetric Signal Analysis

If you are optimizing breakouts, or any other signaling design where there are independent entry-exit parameter settings, you are actually looking at real world signals that appropriately lay across all of the asymmetric tradescape surfaces. You can gain additional confidence in those optimization parameter selections by using asymmetric tradescapes. For example, if you happen to favor HH=10 LL=20 breakout settings, you would want the entire tradescape with an asymmetry of 0.5 to be favorable, regardless of where one's specific signaler happens to rest on the surface. You would also want the effective EM length and the lag fraction of your actual trading signal to rest in a robust region of the tradescape.

Let's look at an asymmetric tradescape study for the traditional turtle breakout settings of HH=55 and LL=20. Such breakouts have a signal asymmetry of 55/20=2.75. We will trade long only and each entity must have 10 years of EOD price history:


This is a selection of US ETFs for various world market entities. We scale the graphs to display only those regions where full accuracy in trading the order produces more reward than pain using the RRt metric. We know the turtle signaling system is capable of relatively low lag fractions, but as with any signal offering these lower lags, it will also suffer a good measure of inaccuracy in terms of trading order. That is expected since breakouts act directly on price, a response to the disorder or chaos in the price movements. For this historical period, certain entities look quite promising. Others do not. A weakly performing entity may be due to the inappropriateness of this slow to enter - fast to exit signaling asymmetry, the lack of growth in the entity across time, or the lack of order in price movements. Keep in mind that the asymmetric analysis, even with the turtle signaling asymmetry, is not the actual turtle breakout-type signaling system. This is a tradescape where there is full accuracy with respect to trading order at every point in the surface and the lag is perfectly uniform (zero scatter) for each entry and exit represented in each of the backtests in the tradescape. Real-world breakouts have neither property. Still, it is instructive to see which entities respond well to trading order with this particular asymmetry.


This is a selection of asymmetric tradescapes for continuous futures using the turtle signaling asymmetry of 2.75. These are 7 year EOD long asymmetric tradescapes. Rather quickly you see one commodity, natural gas (NG), that shows little tradable order. The S&P 500 index futures (ES) appears to be surprisingly good at several time horizons, as does the ED Eurodollar (ED) at fast time horizons and gold (YG,ZG) at intermediate time horizons. Gasoline (RB) shows good trending order at intermediate ranges and crude (CL,QM) works only at very fast or very slow time horizons. The differences between silver (ZI,YI) and gold (ZG,YG) are interesting with gold appearing to be more favorable for intermediate time horizons. Platinum appears to offer strong tradable order at various time horizons. And as one might expect, we see virtually no difference between the full size and the e-minis (CL vs QM, ZG vs YG, ZI vs YI). Keep in mind that these asymmetric tradescapes evaluate a high asymmetry, corresponding with a slow to enter and fast to exit paradigm. The turtles also traded a symmetric HH=20 LL=20 breakout signaling.


These are the 2.75 asymmetry tradescapes for forex pairs. This is for the "long" side, meaning one trades the first entity against the second. For AUDUSD, one is trading the growth of the Australian dollar against the US dollar. This is daily data going back more than ten years. There is some promise for trading the AUDUSD, EURGBP, EURUSD, GBPUSD, and NZDUSD pairs for advantage in the first pair.

In order to test the second side of the pair, we must trade 'short' and we must invert the asymmetry so that the same slow to enter and fast to exit paradigm is adhered to. We thus test a 1/2.75=.36 signal asymmetry.

We do see the three that had no performance on the first entity show some promise on the second entity in the pair. Still, with few exceptions we see that trading the order in currency movements for more reward than pain is difficult. Unlike securities where there is the potential for long term sustained growth across time, or commodities which generally increase with price over time, currency pairs have no intrinsic long term growth for trend trading. While there can be some very long periods where one entity is stronger and the other weaker, the tradescapes essentially show the impact of this theoretical long term net zero growth baseline.

A Signal Analysis Example

Let's look at trading EURUSD long. A ten-year daily data optimization of breakouts produces the following response surface:


If we select the fast region, we have HH=8, LL=16, a signal asymmetry of 0.5.


The blue curve is the underlying and the red is the traded equity curve signaled using this [8,16] breakout. The RRt reward to pain increases from 0.42 to 1.00. The unleveraged robust trend increases from 3.25% to 4.08%. Clearly this is an entity that is very difficult to signal using breakouts, something that tends to be true of most currencies.

So what did the signaler achieve?


The box with the [1] on the plot shows an EM length of 7.38, a lag fraction of 0.79. As one would expect from the [8,16] asymmetric breakout pair, the lags at the entries are much faster than the lags at the exits. The entry lag averaged 4.36 bars and at the exit lag averaged 7.36 bars. The lag used for the plotted point is 5.86. By using the 0.5 asymmetric tradescape as the background, we see the fast breakout optimum at almost precisely the position we would hope to realize. Although the real-world breakout system did not achieve the stronger RRt of the tradescape, it fared quite well for a signal with a 28% coverage error in terms of the turns generated by the ideal asymmetric EM signal.