TSCI.pngPlatform Overview


Platform Editions
Platform Paradigm
Platform Signal Design
Platform Procedures
TradeStation Analysis Techniques

A First Look and Comparing Multiple Securities and Financial Instruments

7901.png The Tradescape... procedure (the TSTradescape analysis technique in the TradeStation interface), is used to get this basic picture of an entity's performance. The Tradescape Platform offers a variety of performance metrics that offer reward-risk and reward-pain estimates. Our preference is for RRt, the platform's default. This estimates reward as a robust CAGR trend in an all-equity-invested trading model and the trading pain as the average retracement from the all-time high.

These procedures are useful for a first picture of what can be expected, at least from a historical perspective, in trading any given instrument and how multiple instruments compare one against the other. To better understand a tradescape, double click on any tradescape point in the surface. You will see the equity curve for the specific backtest represented by this one point. As presently designed, there are 700 backtests in a tradescape.

When multiple panels are shown with Tradescapes, you are looking at the trading landscapes of different entities. The differences can be immense, especially with a daily bar density.

Asymmetric Signaling

An EM signaling algorithm can generate asymmetric signals. This means the time horizon for the entry side of the signaling can be greater or lesser than the exit side. While one would fully expect a fast to enter - slow to exit state during periods of strong positive sentiment toward the entity (and for the market in general), and slow to enter - quick to exit states for the negative sentiment periods, each entity will be different when all of those sentiments are seen in one comprehensive picture.

Those who trade asymmetric signals, such as different breakouts for entries and exits or different MA lengths to generate entries and exits, know they are trading a moving target in terms of sentiment. Still, we feel it is useful to know an entity's behavior in terms of entries and exits over a given period of history. The last thing you may want to do is launch an asymmetric signaling system whose time horizons run squarely against the long term historical behavior.

7902.png The Asymmetric Tradescape... procedure (the TSAsymTradescape analysis technique in the TradeStation interface), is used to get this basic picture of an entity's tilt with respect to signal asymmetry. The Tradescape Platform offers a variety of performance metrics that offer reward-risk and reward-pain estimates. Our preference is for RRt, the platform's default. This estimates reward as a robust CAGR trend in an all-equity-invested trading model and the trading pain as the average retracement from the all-time high.

When multiple panels are shown with Asymmetric Tradescapes, you are usually looking at one entity with many different signaling asymmetries on the EM algorithm. You can also be looking at a favorite or accepted asymmetry with a great many entities.

Stability Across Time

If one has a sufficient price history or trades frequently enough in a shorter measure of history, one can study how the trading landscape changes across time. Entities vary widely. Some are actually robust and have weathered the weak market periods surprisingly well. Others can be fully untradable during certain periods. Further, one may see an optimum trading zone in a comprehensive tradescape, but see that there are time segments where that same zone is highly unfavorable. It is not so much an issue of constancy across time as rather gauging the extent to which the lack of such would break any given trading setup.

This tradescape progression across time is covered in progressive tradescapes.

7903.png The Progressive Tradescape... procedure (the TSProgTradescape analysis technique in the TradeStation interface), is used to get this basic picture of an entity's tilt with respect to signal asymmetry. The Tradescape Platform offers a variety of performance metrics that offer reward-risk and reward-pain estimates. Our preference is for RRt, the platform's default. This estimates reward as a robust CAGR trend in an all-equity-invested trading model and the trading pain as the average retracement from the all-time high.

When multiple panels are shown with Progressive Tradescapes, your are looking at one entity's progression through time.

Sampling Density

Unless one is looking at tic data, each bar's opening and closing prices are snapshots in time, and the high and low represent the scatter in prices across the duration of the bar. There can be profound differences in the tradability one finds with different bars. For each, the order and chaos can look different. Data sampled at 1-min bars responds very quickly to the chaos, which we assume does not distribute as uniformly across time horizons as does order. Reward-pain can look very good at high sample rates, but one may have to trade very frequently in order to realize those benefits. Data sampled end of day filters out much of the local chaos or perturbations, trades can be quite long, but the pain may be far greater. If a trading signaler has an average lag of 20 days, a lot of damage can happen before the signaler exits.

To seek to understand the dynamics of trading different bar data, the platform offers intraday tradescapes. In general, one collects 1min bar data and panels are automatically generated for a variety of bar densities. Even though the process of computing EM models for hundreds of thousands of bars can be extensive, this procedure allows one to see in a few minutes that which would generally require days to generate in conventional backtesting. It is also one of the most effective methods to determine an optimum time horizon to trade as the strongest and most robust time horizons will be clearly evident in the different bar densities which include it.

7911.png The Intraday Tradescape... procedure (the TSIntraTradescape analysis technique in the TradeStation interface, is used to get this picture of order and chaos across the different sampling densities. Faster is not always better even with an ideal execution assumed.

When multiple panels are shown with Intraday Tradescapes, you are looking at one entity's performance across a set of bars. For 1min bar data, progressions through EOD data are automatically generated. If your computer has sufficient memory, you can input a million 1-min points and see ten year's history at every bar density. The EM algorithm trades the gaps, with the corresponding low win rates, although for pure intraday trading with no overnight holds, you can mathematically remove all overnight movements from the data.

Trading Signal Analysis

This is probably the main reason you are using the platform. You want to know how good your trading signal is when processing a given entity. You want to know what you can expect walking forward with a given signaler.

The tradescapes show you where the sweet spot lies in time horizon and bar density and what is realized at various lags, but this is the reality of this gold standard, what is possible from this EM reference set of signals. What you care most about are your real-world signals. Where do they fall on this trading landscape?

7913.png The Trading Signal Analysis... procedure (the TSSigEval analysis technique in the TradeStation interface), is used to analyze your own trading signals. It evaluates one’s own real-world trading signals against the backdrop of an entity’s tradescape. The trading signals for the tradescape are taken directly from the traded entity. The real-world signals are your own and can come from anywhere you choose. The tradescape is the reference against which your signals are measured. The real-world trading signals will be plotted atop the tradescape at their respective time horizon and lag fractions. The signal analysis for each trading system is shown when moving the mouse over the special point representing that particular trading signal. If you double click on a trading system point, you will see the equity curve for its specific backtest and that of the corresponding EM signal that was constructed to match the count of turns in your signal. As this stage, to study the performance of your real-world signaling system across time, it is best to look closely at the actual equity curve and at each trade.

Importantly, you will know to what extent you have captured the tradable order in the price movements. You will know what lag your algorithm achieved at the turns, and the 'effective' information utilization of your algorithm, the corresponding length of the EM algorithm that matches the zero crossings of the differenced signals. Also, you see the coincidence accuracy, the percentage of time your signal turns match those of the EM reference.

In very little time, you will understand why EMAs, SMAs, WMAs, and the countless others designed over time have no clear superior choice in signaling. You will see the extent to which breakouts process both order and chaos. When you elect to trade a signaler, you will know it rests historically in a favorable place. Most things will make sense. You will know, usually without an optimization that accounts being on the right side of trades associated with fat tail events, that the system is about as well designed as it can be taking historical behavior into account. If there has been an real-world optimization, it will be reinforcing to see it rest in a favorable place on the tradescape. If the design is ill-posed, something readily done, this analysis will generally point out the higher risk of implementing such a system.

Two-Stage Signaling (Professional Edition)

The professional version of the platform includes two features that are likely only of interest to financial professionals. The first of these is a two-stage signaling we call sentimentscapes.

One item soon becomes clear when using tradescapes in signal design. The fractal nature of the signaling science is visually apparent. One can perform a two-stage signaling at two different time horizons. The longer time horizon becomes a sentiment windowing procedure. As in the trend filter used in many trading systems, one constructs windows where only long trades and only short trades are permitted using this longer time horizon signal. The shorter time horizon is used to generate the actual entry and exit signals. In the event of long trades, the idea is that periods that have a wide-sense negative sentiment are better off left alone, even if the more local signaler is indicating an entry. In general, one finds a higher win rate from longer time horizons, all else being equal. As the bar density widens, the bar to bar differences become greater relative to the overnight gap.

As things are presently done, accomplishing this in practice with any certainty is next to impossible. There are two signals at two different time horizons. There are two different lags. There is the inaccuracy in placing the long and short trading windows, and inaccuracy in the actual entry and exit signals. In short, there are too many variables to have any effective handle on the process. Further, it is often the case that one wishes to use a different entity for ascertaining the sentiment, such as an overall market index. One then permits long trades only when this sentiment source is in a state of positive sentiment. For the US markets, SPY and QQQ are obvious candidates for a sentiment source. When one adds the task of discovering an effective sentiment source for a given entity, the complexity increases dramatically.

The reason sentimentscapes are so valuable to a professional is that it is often possible, by this two-stage sentiment-augmented signaling, to alter the trading landscape so that far less is needed in terms of heroics in the signaler itself. One may be able to use a basic signaler in this scenario, and it may outperform the very best of direct real-world signalers. In effect, one combines two probabilistic zones, one within the other. The benefits can be significant but likewise so are the challenges. A lot can go wrong when two signals, and possibly two different targets for those signals, need a proper design. In many cases, different real-world algorithms are used for each, as in a moving average crossover for the coarser time horizon and a breakout pair for the finer timer horizon.

7904.png The Sentimentscape... procedure (the TSSentimentscape analysis technique in the TradeStation interface), generates one or more of these sentiment-augmented tradescapes (sentimentscapes).

7905.png The Progressive Sentimentscape... procedure (the TSProgSentimentscape analysis technique in the TradeStation interface), generates a series of sentimentscapes that are sequential in time for a single entity in a chart.

7912.png The Intraday Sentimentscape... procedure (the TSIntraSentimentscape analysis technique in the TradeStation interface), generates a series of intraday sentiment-augmented tradescapes (sentimentscapes) with different bar spacings.

Referential Signaling (Professional Edition)

The second of these professional features involves surrogate signaling. This is often used in basket trading. A basket of securities is traded using just one signal, usually derived from an overall market index or some other instrument representative of the entities in the basket. This approach prevents the whipsaw trades in an individual entity typically caused by specific earnings reports and news. A significant measure of skill is needed to select the securities for such a basket.

To make this process as easy and as painless as possible, analogs exist for all of the above procedures (except for the intraday). In each of these procedures, a surrogate is specified for the signal source. An example would be using QQQ as the signal source for a variety of Nasdaq securities.

There may also be instances where a professional is required to actively trade a given entity whose properties and behaviors little lend themselves to direct signaling (requires a highly demanding signaling algorithm). This is likewise generally a longer term trading scenario, and these referential routines make it much easier to locate a viable surrogate. One then designs the trading system for that surrogate-traded entity pair and a referential tradescape reflects what is possible in this more complex framework.

This is a difficult problem as it is currently addressed, but referential tradescapes make it possible to see in one view how dozens of securities might look with any given market reference as the signaling source. In one view, because of the common structure of a tradescape, it is possible to find all entities that have a favorable time horizon and a forgiving lag tolerance with that reference as the signal source. If one is looking at the Nasdaq 100 securities and QQQ as the signal source, in a few minutes it can be known which entities will trade favorably, where they must be traded in terms of time horizon, and how much lag is tolerated. Even if the time horizons match, if the lag tolerance is missing, the entity probably doesn't belong in the basket.

The very nature of surrogate signaling introduces a certain fuzziness in the tradescape. A surrogate such as SPY or QQQ has its own fuzziness as it can be effectively regarded as an average of many entities. This fuzziness in a referential tradescape is something a trading professional would expect and it is part of the design of such trading systems.

7906.png The Referential Tradescape... procedure (the TSRefTradescape analysis technique in the TradeStation interface) generates one or more referential tradescapes.

7907.png The Referential Asymmetric Tradescape... procedure (the TSRefAsymTradescape analysis technique in the TradeStation interface) generates one or more referential asymmetric tradescapes.

7908.png The Referential Progressive Tradescape... procedure (the TSRefProgTradescape analysis technique in the TradeStation interface) generates a series of referentially signaled tradescapes or asymmetric tradescapes that are sequential in time for a single entity in a chart.

7909.png The Referential Sentimentscape... procedure (the TSRefSentimentscape analysis technique in the TradeStation interface) generates one or more referential sentiment-augmented tradescapes, referential sentimentscapes.

7910.png The Referential Progressive Sentimentscape... procedure (the TSRefProgSentimentscape analysis technique in the TradeStation interface) generates a series of referentially signaled sentimentscapes that are sequential in time for a single entity in a chart.

Advanced Signal Analysis (Professional Edition)

To accommodate the above procedures, there are two professional options where trading signals can be plotted atop referential tradescapes, sentimentscapes, and referential sentimentscapes and where these more complex signals are compared against the corresponding referential and/or two-stage EM reference.

7914.png The Referential Trading Signal Analysis... procedure (the TSRefSigEval analysis technique in the TradeStation interface) evaluates one’s own real-world trading signals against the backdrop of an entity’s referential tradescape. This procedure is used exclusively for single stage surrogate signaling.

7915.png The Advanced Trading Signal Analysis... procedure (the TSAdvSigEval analysis technique in the TradeStation interface) evaluates one’s own real-world trading signals against the backdrop of an entity’s sentimentscape or referential sentimentscape. This procedure is used exclusively for two-stage sentiment-augmented signaling.

Specialty Design Procecures (Professional Edition)

7920.png The Book-Half Tradescape... procedure (the TSBookHalfTradescape analysis technique in the TradeStation interface) automatically generates tradescapes to study the book-half money management strategy.

7921.png The TSBookTimeTradescape... procedure (the TSBookTimeTradescape analysis technique in the TradeStation interface) automatically generates tradescapes where the exit is fixed at a specified count of bars.

Backtest Customization (Professional Edition)

The professional edition also includes the option to alter the settings of the backtest engine used to generate the tradescapes. The defaults are designed to process the EM signal in a mathematical way so as to define a property of the time series. While any real-world factor should be accounted in the actual real-world system being tested, and not in the common framework of the tradescape reference, the professional version does allow certain inefficiencies to be introduced into the tradescape proper. These include market timing for the trades, commissions, slippage and spreads, various types of stops, and the actual entity to be used for the signaling (to accommodate the many algorithms that use HLC and HLC values instead of C to generate the signals).

Because these issues can impact the tradescape's surface, and in some instances, the signal itself and its time horizon and accuracy, we limit these adjustments to those who are mostly likely to understand the implications and many subtleties that could invalidate an analysis. The choice to trade HLC instead of C, for example, might be small, but stops that become a part of a trading system are anything but minor. It can sometimes be instructive to see what an alternate signal price or stops do to the EM signal which can be assumed to have close to full accuracy. If a factor adversely impacts the EM signal, it should be carefully studied with one's real-world signal before implementing. Each signal analysis call can have up to fifteen signals on the tradescape. Each of the variations should be explored with a real-world signal. No system component, such as an HL variable or a trailing stop should be implemented in a real world system based upon the behavior of the EM signals. This is another reason this option is limited to the professional edition.

8617.png The Customize Backtest Engine option in the File menu allows optional settings for the backtest engine to accommodate market timing, costs, signaling, stops. etc.