Trading Platforms | Globus Trading

Sterling Trader PRO

Globus Trader PRO

Laser

Globus Trade Partner

Globus Trade Partner is a software package that allows traders to build, test, optimize and run various trading strategies - no programming required.

Trade Partner consists of two applications - Strategy Builder and Strategy Runner. The Builder handles all of the analytical tasks and the Runner executes the strategies prepared by the Builder.

Product Features

Globus Strategy Builder

Visual script builder that allows traders design their trading strategies from a large selection of building blocks
Accurate trade evaluation using individual tick data. Tick data are available for 2,600 symbols
Sophisticated trade management including variety of trailing stops and position management options
Multiple order handling options for position entry and exit
Strategy parameter optimization using Particle Swarm Optimization method that efficiently finds results for practically any number of parameter combinations
Stock screener function that allows traders interactively pick the list of symbols using multiple fundamental, statistical and technical analysis criteria
Ability to automatically scan large number of symbols using multiple strategies, optimize individually each strategy-symbol combination and verify results using walk-forward test
Multiple-core CPU support significantly improves optimization performance
Hierarchical system of user access rights gives managers ability to control strategies developed by their traders and also protects strategy content from unauthorized access

Globus Strategy Runner

Run large number of strategies and symbols from a single computer
Run multiple strategies on the same symbol
Supports large number of position entry, exit and management policies
Allows the trader to execute independent manual trades on the same symbol/account
Can be safely restarted in the case of Internet connectivity disruptions

Note: Order execution requires Sterling Trader Pro software and account. Direct order execution is currently only offered to the traders who have accounts with Globus Trading LLC.

System Requirements

Windows XP, Vista or 7. 32 and 64-bit versions are supported
2 GB of RAM (4GB recommended)
Multi-core CPU recommended

Using The Software - Quick Reference

Globus Trade Partner implements the whole cycle of strategy development and execution. This cycle consists of the following steps

Strategy creation and backtesting where you put your ideas into a strategy and run it against historical data to verify its viability
Strategy optimization where you find the optimal combination of parameters that provide the best risk/reward ratio for selected historical period
Stock screening where you select the symbols that meet your criteria and are good candidates for your strategy
Walkforward test, where you compare performance of the strategy over the optimization period to its performance on the “out-of-sample” data, i.e. the data that were not used during the optimization stage
Strategy execution, where you run your strategies

In a typical scenario strategy developers first build and refine their strategy and then run it repeatedly through the screen-optimize-walkforward-execute cycle. Behaviour of the stocks tends to change over time, so the strategy that worked really well in the past is likely to go out of sync after some time and stop making money.

Strategy Creation and Backtesting

To build your first strategy, start Globus Strategy Builder and log in. Once the application starts, go to the first tab on the interface. There are a few sample strategies included with the software, so the best way to understand how the software works is to try one of them before building your own.

The left part of the interface lists all of the strategy parameters and smaller pane at the bottom left corner explains the meaning of the selected parameter. Some of the parameters, like Symbol or Position Size are directly editable, others, like trailing stop pop-up additional window where various options can be selected.

To run a backtest on the strategy, click the Run button in the bottom left corner. The strategy will be evaluated using both minute bar values and tick data (individual sales and quotes during the day), so when you run the strategy on a new symbol it will take a little bit of time to download the data. After that, the data are cached on your hard drive in a highly compressed format, so the next run on the same symbol will be executed much quicker.

Once the backtest is complete, you will see the chart with all the entries marked by icons and lines. Short signals are indicated by the arrows pointing down from the top, while the long signals are shown by the arrows pointing up. Position exits are marked by the icons showing whether transaction ended up in a gain or a loss. The panel at the bottom provides various strategy performance metrics. The next tab in that panel lists all the signals with their time and execution prices. Double clicking on the signal will move the chart so that the signal you clicked is positioned in the center.

The buttons at the top of the chart provide a number of convenience functions, including

Series window that shows the values of price and volume bars as well as any indicator displayed on the chart
Quotes window that shows individual tick data for selected bar to help the user understand how the entry / exit point was evaluated
Indicators windows comes in handy for more complex strategies that use large number of technical analysis indicators. By default all of the indicators and values used in the strategy are displayed on the chart, and indicator window gives user ability to show/hide them individually or in groups by condition type (for example short entry)
Conditions window has similar function, but instead of listing individual indicators it shows all of the conditions used by the strategy and gives user ability to show and hide them at will

The line of red and green circles underneath the chart (better visible at higher zoom) shows which of the conditions are valid on each of the bars. Our sample strategy has only one condition, so there is only one line. More complex strategies with multiple conditions will show each of those conditions on a separate line making it easier to figure out whether the strategy works as designed.

Once you run the sample strategy you will notice that it performs poorly and loses money. To improve the situation move to the second tab of the main interface called “Optimize”. The left panel of Optimizer looks similar to the strategy parameters panel in the Backtest tab. The only difference is that any numerical parameter of the strategy allows you to specify the range. The job of the Optimizer is to find the combination of parameters values from the user defined ranges that delivers the best strategy perfromance. If you click on the Summary button in the bottom left corner you will see the list of all parameters with their ranges.

Even for this simple strategy the number of possible parameter combinations is in the millions. For more complex strategies the number quickly becomes astronomical. Going through all possible combinations of all parameters is simply not viable. The task is further complicated by the fact that parameters has strong cross-dependency. This means that you cannot, for example, first optimize values of number of periods used in the strategy technical analysis indicators and then optimize profit target / stop loss.

In order to solve this problem we have implemented Particle Swarm Optimization algorithm and adapted it to this particular task. This algorithm was first published in 1995 and has gained popularity in financial industry since. This algorithm uses a large number of particles that are initially scattered randomly in the parameter space. Once the particles are distributed they start looking for local maximums of the cost function and also communicate their findings to each other. As a result they do not get stuck on local peaks, but instead gradually converge on the most prominent peak.

To start optimization click Start button at the bottom. The pop-up window will allow you to select optimization parameters. The most important parameter is the type of cost function. Optimizer decides which combination of parameters is better by comparing the values of cost function resulting from those parameter combinations. It is possible to make Optimizer search for maximum overall profit or profit per trade, but those results will be only taking into account one particular type of reward without any reagrd to risk. We recommend using combined cost function that takes into account several parameters at once, including net profit, maximum drawdown, profit per trade and time in position. This function allows the algorithm to select the value that strategy developer would have selected using his or her best judgement of risk and reward.

Other algorithm parameters are for more advanced users to choose. The filters allow users to prevent the algorithm from finding solutions that do not meet criteria set by the user, for example the limit on maximum drawdown. When those limitations are known, the computational resources are spend more efficiently in the area that meets those criteria.

For the first run we recommend to stick to default values. Once you hit Start button, you will see the graphical window that shows the progress of optimization. Each of the parameters being optimized is assigned a colour and the values of cost function for different parameter values are shown with coloured dots. As the optimization goes forward you will see that the dots will start migrating toward the top of the screen, showing improvement in the value of cost function. The pane on the right shows current best set of parameters and the corresponding value of cost function. The combined cost function uses logarithmic scale, and therefore does not grow too much. Other costs functions (like net profit) are displayed in their natural scale.

Once the process converges on the solution (i.e. subsequent iterations stop improving the value of the cost function) the Optimizer performs sensitivity analysis of the solution. It dissects the discovered peak and shows its cross-section by each of the parameters on the same screen. It also measures the quality of the peak. Sharp peaks signify solutions that are likely only a result of a curve fitting and have little chance of success on out of sample data. Therefore, the sharper the peak the lower the quality score. Overall peak quality score value is calculated by combining quality scores for each of the peak's dimensions.

The user has an option to make the Optimizer to search for several peaks in different areas of the parameters space. For that purpose the area around the first peak is “cordoned off” and the particles continue their search for another, distinct peak. The user is thus given the choice to either apply his or her judgement to select the better of the discovered solutions or spread the risk by distributing share lot allocated to trade this strategy between several distinct versions of it.

Once the process is completed you can see 100 best and 100 worst results in the Results tab. They are sorted by the value of the cost function, but have all of the statistics to show the user how the decision was made. As you can see the improvement for our sample strategy is quite dramatic – instead of losing money it now makes good returns. At the same time it should be noted that overall peak quality score for this strategy is very low because some of the parameters produce extremely sharp peaks and even the smallest change in those parameters decimates returns. This demonstration shows you that Optimizer can find profitable solution for this strategy, but but at the same time it provides the quality metrics that show that this strategy is not a good candidate for trading in the real world.

This optimization process can be easily repeated for a large number of symbols and strategies. For that purpose you can select the symbols that meet the variety of criteria using interactive Stock Screener. Once you trim down the selection to a reasonable number you can save this filter for later use. It should be noted that you are saving the filter, no the results that it generates, so your scan will never become stale. For example, if you were looking for stocks that trend up for 7 days, the same filter will give you updated results every day.

Once you have your symbol selection figured out, you can move to the next tab called Walk Forward. There you will see the list of all your strategies and a few configuration parameters. The purpose of the walk forward test is to find out which strategies perform well on out-of-sample data. For that purpose optimization is performed using the Backward number of calendar days. Then the parameters optimized for that period are applied to the Forward number of calendar days. The results table on the right shows the performance of the strategy over both time periods on adjacent lines. By comparing performance metrics between opt leaving user ability to select the strategies that fare well either manually or using the filter function.

Walk forward function allows the user to process large number of stocks and strategies in a fully automated fashion. For example, run the scan overnight and pick the strategies to run in the morning. Given that those calculations run in unattended mode, the optimization graph Is not shown. Instead all of the available cores are used for calculations. The test have shown that on a 4-core machine walk forward test runs 3.5 times faster than then the same optimization task for a single stock with graph enabled.

Now that you have gotten the taste of those tools it is time to design your first strategy. To start you need to move back to Backtest tab, open File menu at the top and select New Strategy.