button to add all variables, including the first. These release notes list feature changes for the current release of Stella Architect, iThink Version 2, Stella Professional, and Stella Designer. Feature changes are also shown for previous versions of the software. There is a list of Known Issues for the current software release as well as a version by version detailing of Defects Fixed.
It is now possible to load in data as a run in Stella Simulator. This is most useful for calibration runs as described in Stella Simulator Use Examples.
When using the builtin SIZE to provide the number of elements in an array dimension you can now simply specify the dimension name. See Array builtins for details.
The translation utility has been updated to handle more Vensim models, especially those that use flows as inflows to (or outflows from) than one stock equation. See Opening Other Models for more discussion.
It is now possible to use equations to specify the internal content of conveyors and queues, as well as to specify the initial values of attributes and cycle time within stocks. These additions allow a model to be initialized in a way that is representative of the true initial state of the system (for example a conveyor with initial growth). For more information on this see Initializing Discrete Stocks.
To simplify setting optional equations rather than having to turn on and off specific optional equations, such as that to arrest a conveyor, the equations can now simply be left blank. See Equation Tab for more discussion.
The interface used to find variables has been streamlined to make it easier to perform the most common tasks. See Find window for more details.
The right click menu can now be used to show the causes and uses of a variable. In addition, the causes and uses now list both first and second order causes and uses, making it easier to visual dependencies. See Context Menu and Causes Uses for discussion.
You can set up multiplayer games to automatically advance based on a timer. This forces games to continue even when some players have not met the deadline of committing their decisions. See Multiplayer Options dialog for details.
Files saved will no longer change when there have been no changes to the underlying model. This used to happen because of the sequence in which content is written out and changes in internal object IDs.) The invariance includes modules saved as separate files which facilitates use with version control systems. See Saving Changes to Modules for information of saving modules as separate files.
The "Report as" option associated with a converter will now be used on the interface to determine whether the current value of the variable will display when a model is paused. Selecting stock will ensure that the current value is always displayed (even if it might change after resuming). See Scales and Ranges Tab for details on setting this.
An option has been added to allow diagrams to draw only the outline of shapes. This is useful when you are using a diagram to overlay other content. See the Model View documentation for details.
If you draw a connector that already exists you will have the option to replace the old connector with the new. This is helpful for moving connections to ghosts, and also rearranging connectors in busy models.
You can now use Ctrl+E to switch to edit Model and Shift+Ctrl+E to switch to explore or experiment mode. A full list of keyboard shortcuts is available at Keyboard Shortcuts and Key Combinations.
The new builtin ALLOCATE allows allocation of quantities taking into account both availability and attractiveness. This can be used to allocate a supply to different consumers, or to allocate demand to different producers. See ALLOCATE(<available>, <index>, <target_array>, <priority_array>, <priority_spread>, [<shape>])
for more discussion.
You can now specify a precision for graphical inputs on the interface allowing, for example, to keep to integer values. See Graphical input (Interface) for detail.
It is now easy to find every variable that uses a given dimension. This is useful for when dimensions are changed, and also for finding unused dimensions. See Find window for details.
Loops that cross modules will now highlight the modules involved in the loops and the associated module connectors (if they represent only a single cross level connection). This makes is somewhat easier to follow loops across modules, but the automated CLD view (see CLD Generation) may be a more effective way of visualizing such loops.
You can include html tags in text boxes and annotation on the interface. These will display and behave the way they would in a web page, and can provide a convenient way to add links (or embed content) in interfaces. See Text boxes and Annotation Button (Interface).
Understanding models has never been easier! Loops that Matter is new functionality in Stella 2 that allows you to immediately visualize the impact of every variable in a model on every other variable during the course of the simulation. The software identifies the feedback loops important to generating behavior and allows you to create causal loop diagrams that display those loops in a simplified form. For more see the Loops That MatterTM Overview.
Stella will now create stock and flow diagrams, or fill in missing pieces, for models created in other applications and brought in to Stella. See Opening Other Models for more details.
When a flow is forced to 0 because of a non-negative stock, both the flow and the stock will now highlight to show that this has happened. Hovering over the highlighted stock or flow will give a message showing what type of constraint was encountered. There is an option in the Preferences Dialog to turn off unless an option in the Model Settings Properties Panel is used to turn it off for a specific model.
New to Stella 2.0 is the ability to open Vensim .mdl files directly. Both the equations and diagram will be translated. See Opening Other Models for more details.
The Equation Viewer will now display equations in a table, making it much easier to read and navigate through the listing. The table can be saved to html or copied to the clipboard in for pasting into a spreadsheet.
The LOOKUPMEAN builtin can be used to compute the mean value of a graphical. This is most useful for graphicals on the range of 0 to 1 representing probability density functions. The builtin MLPANN can be used to embed artificial neural nets in models to capture arbitrary nonlinear relationships. The builtins PATHSCORE and LOOPSCORE are for use with Loops that Matter. See Miscellaneous builtins for more details.
Material coming into a conveyor is normally put at the beginning, then sent across the conveyor over the course of the simulation. It is now possible to distribute the material over the conveyor following the contour in a graphical. This effectively allows arbitrary residence time distributions which provides more accurate modeling in cases where the time in stock is variable. See Spreading Conveyor Inputs for more detail on how this works.
We have added the ability to create an interface where Stella Live is always on. Simply set this option in the Interface Settings Properties Panel. There is also a new action to first run a model, and then enter Stella Live, mimicking the behavior of the run button during model development (see Actions (Interface)).
You can now save an optimization log to disk during the optimization. This allows recovery in case of a software failure which is valuable for optimizations that run for long periods of time. See Optimizer Messages for more detail.
During data collections the results from the last Stella Live run can now be captured (see Data Collection for more details).
It is now possible to create and use subranges to simplify both array operations and the selection of values from arrays with different dimensions. See Subranges for more details.
You can now set the size of CLD pages making it easier to organize your content. See CLD Style Properties Panel for details.
We have added an option to include precomputed converters and flows when exporting data. See Export Data dialog box for more details.
We have added a function that allows you to quickly compare two runs. This makes it easy to track down differences between runs, especially differences resulting from changes in constant or graphical values. The Runs Comparison window reports the differences between two runs and the first time those differences occur. It is opened from the Open Runs Comparison command on the Windows menu. See Runs Comparison for more details.
You can now control the speed at which a simulation progresses using a variable(rather than the old simulation speed control). Simply select the model variable you want to define the simulation duration with in the Model Settings Properties Panel. You can use a slider or other input device the then set the simulation duration or speed. You can also vary the simulation speed based on model conditions.
There is an additional export option that reads a list of variables from a file and then exports values for those variables. This is most useful for batch processing where the list of values to view is not known ahead of time and can be written to a file allowing a model to report different results without having to edit the model. See Export Data dialog box for details.
It is now possible to create definitions for units that already have built in definitions, effectively replacing them. This makes it easier to have units suited to your needs. See Unit Editor dialog box for details. This dialog has also been improved so that double clicking is not required to edit unit definitions. In additions, changes to unit definitions are now tracked in the undo/redo system. We have also added Tons and Tonnes to the built in definitions.
Causes, Uses, and Selection Hisotry
The selection history panel has been replaced with menu items in the Equation Tab of the Properties Panel and also on the Results Panel. These menu items allow you to look at the previous 10 variables you have worked with and, if you select a previous variables, the next ones that you worked with. See
You can now override the settings for your computer to choose whether a , or . should be used as the decimal delimiter. This is done from the Preferences Dialog.
When dragging things on the model dragging off the current boundary of the model will no longer aromatically add pages. Instead the page will be added only after you release the mouse button. Thus, moving off then back onto the current model extent will not cause a change in the number of pages used for the model diagram.
You can now set the border radius on text boxes (Text Box Properties Panel) and annotation popups (Annotation Button (Interface)).
There is now an option to add a grid line at 0 on graphs. See Graph Settings Properties Panel.
Queue outflows can assign a list of attribute values that they will dispatch. This allows downstream routing of material flows based on attributes without having to create separate flow channels for common processing resources. See Queue Computation and Equation Tab for more discussion.
You can now use expressions such as SELF[Dimension-1] inside of equations to reference prior (or later) elements in arrays. See SELF for an example. Similarly, you can create loops with apparent circularity that resolve on an element by element basis. See Referencing other Array Elements for more discussion.
We have generalized flow concept to allow you to specify that a converter should be reported the same way a stock is on tables. This is most useful on the interface, where stock values are reported at the current time when a model is paused, but converter values are not. It is also helpful on financial tables, where converter values are reported one DT in advance of the reported time for stocks (the end of period concept). See Scales and Ranges Tab and Financial Tables for details on setting these.
There is now a number formatting button on the panel used to set up tables and financial tables, making it easier to adjust these to your needs. See Table Properties and Financial Tables for usage.
The simulation log only opens when model opens if there are errors to report. Parameter overrides are still reported, but will not cause the log to open. There is also a new shortcut on the Run Toolbar to open the Message Log whenever it has content.
There is a new builtin IRR that computes the internal rate of return on a stream of payments. See Financial builtins
for details.
The Cycle-time builtins now take an optional attribute range which allows you to look at the characteristics of batches with specific attributes. See Cycle-time builtins for more detail.
The causal lens allows you to (and by default does) filter out causes from initialization as well as constants when looking at the causes. This removes variables that do not directly contribute to the dynamic behavior of another variable. See Causal Lens™ for details.
There is now a FIFO option on conveyors so that they maintain material in the order received and respond immediately to changes in transit time. See FIFO Conveyor Computation for more discussion. Conveyors are marked FIFO in the Conveyor parameters.
You can now set an attribute on flows that are connected to discrete stocks (and non-negative stocks in Cycle-Time Mode). The attributes can be measured on downstream stocks and flows and can optionally be used to prioritize material for processing in queues. See Attribute and Cycle Time Tracking for an overview, Flow options for a description of how to set the attribute, and Discrete builtins for the new builtins used to measure the attributes.
The discrete attribute can be used to prioritize material entering the queue (so the queue is not longer FIFO). This allows material from different sources to be processed differently. See Equation Tabfor the use of the attributes to prioritize material flowing into queues and Queue Computation for more discussion of how the material is ordered.
Queue dispatching to multiple places now allows you to set priorities for destinations and will provide a round robin mechanism to make use of all available destinations (typically ovens). See Equation Tabto set the parameters and Queue Computation for more discussion of how the priorities are used.
Material that has been in a queue for too long can now be purged from the queue through an outflow that is only active when the age of material has exceeded a threshold. See Flow options to set a flow so that it will purge a queue and Queue Computation for more discussion of the way purging works.
Conveyor inflows can now be marked to be spread across the existing material in a conveyor rather than added at the beginning of a conveyor. This is useful when parallel conveyors are tracking similar groups of things, such as aging people, and one conveyor leakage can be used to introduce material to another as it would with incidence of a disease. It is also useful in demographic modeling in general as it makes is much easier to deal with immigration and emigration. See Flow options to set the flow and Spreading Conveyor Inputs for more detail on how this works.
Graphical functions now have a number of suggested forms available that make it easier to draw smooth curves. See Graphical Function Tab for more discussion.
Two new builtins have been added to make it easier to work with graphical functions. LOOKUPINV will invert a graphical so that, if the graphical is monotonic, the value of the input required to get the specified output will be returned. LOOKUPAREA will return the area underneath the graphical up until the indicated input value. For more details see Miscellaneous builtins
The ROOTN builtin returns the nth root of the first argument. See Mathematical builtins for details.
You can now define the elements of an array (both their names and the number of them) using Import Sheets. This allows you to change the specific content of array elements without having to edit the model, and can be very useful for data intensive model set up. See Import Data dialog box for details.
We have added a new graph form that allows you to represent two sets of equally sized array entries on the left and right hand side of mirrored horizontal bar graphs . A common use of this is a population pyramid which shows the number of people by age group separated by sex. See Pyramid Graph for more discussion.
The bar graph now has a variation in which the bars are drawn horizontally instead of vertically. With the exception of orientation, the options and behavior of the two types of bar graphs is identical. For more information see Bar Graph
A new financial reporting style table has been added on the interface to make it easier to present information in the style of financial reports. It allows formatting, including font and color, to be applied line by line. For more information see Financial Tables.
When you place graphs on the interface you now have an option to provide an expansion button for the graph. This will make the graph bigger, and also provide a control so that the interface user can zoom in to see a narrower time range on the graph. See Expanded Graph dialog (interface only)
Published interfaces can now support keyboard input and alternative text to improve accessibility. See Interface Accessibility for more discussion.
You can now add an entire array (or a slice of an array) to a graph and have it stay in the graph in that form. This simplifies graph definitions and improves labeling while also supporting changes to array element definitions without having to rebuild a graph. See Graph Series Property Panel for more discussion.
We have made a variety of usability improvements that should make it easier to build and use models including:
button to add all variables, including the first. There is a new style of the slider (horizontal only) that can give your interface a more modern look. See Slider (Interface)
There is now a panel that will open and list all spelling errors that have been found in variable names. See Spell Check Panel for more details. Improvements have also been made to the ignore selection in spell check so that anything ignored will be remembered with the model.
The module navigation panel allows you to see all the modules in a model and their containment structure in one place. It is a very useful device for quickly navigating through large models. See Module Navigation Panel for more details.
You can now put labels on modules so that the module name used in graph, tables, and cross level references can be short, while the name displayed on the module can be more descriptive. See the Module Properties Tab for more discussion.
The file menu now has a function to export your model as a bundle. This creates a zip file including your model and all supporting files (images, movies, import sources and export destinations). For more details see File Menu.
If you specify a lead-in time for use in an interface then results will always be shown till the lead in time in presentation mode and when published (no blank graphs). See Interface Settings Properties Panel-to see how to set the lead-in time.
A new option to lock a model diagram can be used to prevent any movement of variables or other objects in the model. This is useful for preventing inadvertent changes or reminding others that things should not be changed. This is set in the Model Settings Properties Panel
Allows you to imbed an iframe into an interface page supporting surveys and inclusion of content from other websites on the interface. See iframe(interface) for more information.
The messages related to units issues for unit converting flows have been improved. Checking has also been added for conveyor transit times, conveyor capacity limits, conveyor inflow limits, and oven cook times.
Selecting a text color close to white no longer requires that you set the background in the editing window to see the text. Instead the lack of contrast is detected and a different background used automatically. See Text Box Properties Panel for more about Text Boxes.
The legends now use a square instead of a line segment to show what is being displayed. See Scatter Graph
and Bubble Chart for details on these two graph types.
You can now see results for arrayed variables and adjust inputs via knobs when using Stella live. The array elements being displayed or changed are determined by the selection you make in the Results Panel. When you open a model the first elements of each array will be selected.
By using array wildcards you can set up graphs, tables, and other output and input objects to display different elements of an array based on user selections. On the model layer this works very much like Stella Live with arrayed variables, on the interface it provides flexibility in page design and can also be used to simplify multiplayer game design as described in Using Wildcards for Multiplayer Games. For more information see Wildcards.
Place holder variable allow you to create graphs, tables, and other input and output objects that can be used with different variables based on selections you or an interface user makes. This makes it possible to reuse parts of an interface page, or an entire page, to represent different parts of a model. It also allows the creation of graphs, tables, and numeric displays on models that can be switched by changing which placeholder variable is selected via the Results Panel. For more information see Placeholders.
Confidence bounds allow you to see how much an optimization parameter can change in order to have a specific influence on the resulting payoff. Most commonly they are used to bound parameter estimates resulting from a calibration, but they can also be used in outcome optimizations. For more information see Calibration.
We have improved the reporting of information from optimization by writing it to the Optimizer Messages which allows you to see more clearly the results, and also to look back at previous optimization runs in an organized manner.
We have added the ability to export the set of variables that have been changed on the interface, both constant values and graphicals. This makes it possible to save the settings for a run made on the interface and then reload them at a later time. See Actions (Interface) for details.
We have improved simulation speed and model save time for large models with complex equations.
Team and player assignments allow you to set up classrooms with individuals assigned to specific teams for a multiplayer game. See Team and Role Assignment for more discussion.
This new builtin allows tracking of the run number within a sensitivity run. This is useful if you want to set something specific, such as a noise seed, based on which sensitivity run is occurring. See Miscellaneous builtins for details.
The SELF builtin can now be used in the ENDVAL builtin, allowing customized definitions of RUNCOUNT. See Data builtins
for more discussion.
The Scatter Plot and Bubble Chart now allow you to specify the time range from which values should be drawn. See Graph Settings Properties Panel
Calibration is the process of adjusting model parameters to make the behavior line up with a target, most commonly measured time series data. It is now easy to set up payoffs and use the optimization engine in Stella to accomplish this. For more details see Calibration for more discussion.
Time varying values allow you to have variables take on different values at different times based on imported data. This is useful as a mechanism for bringing in a sequence of decisions or controlling values, and also loading in data to be used in calibration. See Import Data dialog box for more details and Defining Calibration Payoffs
for the use of imported time varying data in payoff definitions.
The -ro option has been added to the command line to run optimizations. See Running Models Using the Command Line
for details.
The same option has also been added to Stella Simulator, as has support for sensitivity and more control over initialization. See Stella Simulator Arguments for more details.
Print and Print Preview have been added to the File menu on the interface window. These will print the pages currently displayed in the Page Preview Panel. See File Menu
There is an additional option to use Sobol sequences when preforming multivariate sensitivity. Sobol sequences are designed to maximize the exploration of multidimensional parameter spaces. See Sensitivity Specs Panel for more details.
It is now possible to replace a run with another run so that any graphs or tables referring to the run will be updated. For example, if you have a run named base, you can have the current run replace it. For details see Data Manager
You can now specify the units of measure for a User Macro based on the units of measure of the arguments to the macro allowing full checking of model units. See Working with Macros for more discussion.
There is a new function INTERPOLATE that will take an N dimensional set of points and return the value derived by interpolating between these point values. In one dimension this will act the same way that a graphical would on a vector of numbers. In higher dimensions the interpolation is between planes (or hyperplanes) of the different dimensions. See Array builtins for more details.
You can now set runs to start part way into the simulation when working on the interface. This is convenient if you want users to start with some results already available. There is also an option to restore inputs when starting new runs, which is very useful for multiplayer gaming. See Interface Settings Properties Panel for more details.
The shape tool makes it easy to add vertical and horizontal separators on interface pages as well as to include outlined or shaded ellipses (including circles) and rectangles. See Shape (interface) for more discussion.
There is now an option on Graphical Inputs to display the current x and y values as the user drags the points in the shape. This is especially useful for discrete graphicals. See Graphical input (Interface) for more discussion.
You can now control when hover tips appear on sliders (always, never, or when the user is dragging the slider). See Slider (Interface) for details.
You can now specify number format for the numeric display in the properties panel. This allows you select multiple numeric displays and set the formatting options for all of them together. See Numeric Display Properties
If you include Loop Indicators on your model they will now be active when the model structure is presented in either a model view or on a story in the interface. See Loop Labels for more discussion.
Causal loop diagrams created in The CLD Window will now have NAN equations automatically added. These will simplify the inclusion of CLDs in simulating models as they will no longer create invalid equation. The NAN(a,b) format can be added by hand when working in regular models (see Creating CLDs) .
Stella will now work with the features in the newly released Mojave operating system on the Macintosh including the dark mode interface selection.
When you replace the contents of a module using the Import button any cross level connections between the module and the rest of the model will be retained as long as the names in the model for inputs and outputs have not changed. See Module Properties Tab for more discussion or Agent Based Examples for a typical usage.
You can now ask for a new causal loop diagram and have the simplified toolset necessary for creating causal loop diagrams. That toolset (
) makes the creation and editing of CLDs quick and easy. It also matches the toolset of the Online CLD Editor. For more information see
The CLD Window.
You can now set the color that a switch on the interface displays when it is on. See Switch (Interface) for details.
You can now set the line color for the graphical input on the interface and also specify a fill intensity to display color below the line. See Graphical input (Interface) for more detail.
The Allocation Input object now has the option to computer only integer values allowing more control of both the values and the label appearance. See Allocation Input (Interface) for more details.
There is now a menu option to download the help files as a zipped archive for use offline. This allows you to view the help files locally when you do not have an internet connection. See Help Menu for details.
We have added to the options for sharing Models and CLDs online allowing you to give other people access to your work without marking it as creative commons. This option is not available in the free online editor. See Sharing Models Online for more discussion.
If there are long variable names in the causal lens, you can now hover over them to see the full expanded name. See Causal Lens™ for more discussion of the Causal Lens.
When a conveyor has multiple leakages with different leak zones you can now specify whether the leakage fraction on later leak zones applies to the inflowing quantity or the quantity that has remaining after the earlier leakages. See Traditional Conveyor Computation for more discussion.
The Interface has a new tool used to add a set of navigation links to an interface page or template. The bar of navigation links can appear as a row of tabs or buttons. The styling includes radius corners and spacing. For more details see Navigation Control (Interface).
Spell check is now available for variable names, documentation, and labels. You can select different dictionaries in the Preferences Dialog and correct misspellings using the Context Menu.
It is now possible to turn modules on and off in Stella Live, while paused, and on the interface, including in storytelling (see Running Modules for more discussion).
Leakage computations on conveyors with a variable transit time have been improved to more accurately reflect the leak fraction or rate. See Traditional Conveyor Computation for a more detailed discussion of leakages and conveyor initialization and computation.
Encapsulation allows you to take a selection of variables, or a sector, and automatically put them into a module with the appropriate cross level connections into and out of it. This makes it easy to simplify models that have too much detail built up in any given module. For more details see Encapsulating Structure.
You can now create a container that includes both graphs and tables making it easier to provide the output you want without using up real estate. See Graph Series Property Panel and Table Properties for more details.
Large models should now open more quickly.
You can now set interfaces to scale prepositionally when they are deployed. This makes it easier to have interfaces that work on different devices without having to limit the minimum or maximum size (another added scaling capability). See Resizable Interfaces and the Resizable Options dialog for more information.
You can now create reusable pieces of model structure that can be used the same way that the builtin functions are. This allows you to create customized versions of commonly occurring patterns such as cascaded delays as well as more specialized functions. See Working with Macros for more details.
Bubble charts allow you to display three different dimensions of attributes on a single graph. These graphs represent a snapshot of values at a time and will animate during simulation or when dragging the time slider. See Bubble Chart
for more detail.
Scatter charts can now have more than a single entry. This allows for comparison of x and y values across different runs or comparison of different sets of x and y values. See Scatter Graph for more discussion.
It is now possible to import data with arbitrary time values rather than requiring a uniformly increasing time axis. This makes it easier to accommodate missing values as well as values from different data sources. See Data Manager for more detail.
Stashable graphs and tables allow you to hold on to run results that are of special interest, while displaying the current run at all times. This is a convenient way to collect a sequence of interesting outcomes, and replaces the old functionality related to persevering and releasing data. For more information see Stash Data for more discussion.
It is now possible to peek into modules without navigating into them. This allows you to see what is happening in a module while still looking at what is happening on the parent model of that module. This allows you to make changes in Stella Live and see the consequences of those changes in another module. It also allows you to get an overview of module structure without losing your current visual focus. See Picture in Picture Module Viewing for more details.
Many of the panels now include sections that can be collapsed if you do not need to refer to them. This decreases the amount of information displayed on the panel and thus easier to work with. See Stella Modes, Views, and Panels
for more details.
A new feature has been added that allows you to compare the model you currently have open with another saved model. This is a convenient way to see what changed have been made for models derived from one another or a common source. See Model Comparison dialog box for details.
You can now zoom in on the displayed time in the Results Panel and Causal Lens. This makes it easier to trace down sources of behavior at a point in time, as well as seeing changes in behavior between multiple runs. See Results Panel
and Causal Lens™ for details.
We have added options to the equation viewer to make it easier to provide written model documentation to meet the requirements of peer reviewed journals such as the System Dynamics Review. See Equation Viewer for more details.
There is now an option to display sliders vertically as well as horizontally. See Slider (Interface) for more details.
You can now specify that graph be drawn without connecting the points into line. See Graph Series Property Panel for more detail.
There is a new option to have buttons change their appearance as the mouse cursor passes over them. See Styles Tab for details.
When you select 3 or more objects on the interface there is now a menu option to either align or distribute those objects to clean up layout. Details are available in Context Menu.
You can keep the movement of objects completely horizontal or completely vertical by holding down the Shift key. This works on both the model and the interface.
The loop indicator now works by selecting all of the elements you want highlighted. This allows you to select more specific loops or parts of structure for display. See Loop Labels.
The find will now retain what you are searching (Variable Name, Equation or Documentation), the order in which to display results, and whether to search in the whole model, the current module or the current module and modules it contains.
You can now provide an external list of parameters to change when running sensitivity analysis, and we have changed the sampling for Ad-Hoc and Incremental when running a limited number of runs to better support random selection of discrete values. See Sensitivity panel for more details.
We have made a number of internal improvements to the security settings (see Model Security dialog box) and fixed issues that arose from specific character combinations in low security passwords. One consequence of this is that models saved in the current version using security settings will not open in versions before version 1.7.
In addition to being able to run optimizations over sensitivity runs you can now run an optimization within each sensitivity run. This is a convenient way to see how the optimal parameters found vary based on other model parameters or noise seeds. This is set up in the Sensitivity Specs Panel
It is now possible to add constraints to differential evolution optimizations allowing the specification of inadmissible search spaces without degrading the efficiency of the search. See Optimization Specs for details on setting this up.
The find tool now makes it easier to look for content that appears in equations and documentation by allowing you to specify search terms within these. See Find window for more details.
A Manage option was added to the right click (Ctrl-click/⌘-click) menu item for formatting items in graphs and tables. This will open the Manage Time Labels dialog box allowing you to add and modify formats as well as changing the default time format.
The scroll position on the table in the Results Panel is now retained when moving between variables.
Options have been added to the Equation Viewer to make it easier to create model documentation that meets your needs. It will now display optional variable usage and provides increased control over annotation. See Equation Viewer
for more information.
Sketchables with Fill
You can now fill in the area underneath a sketchable graph. This is especially useful when using discrete graphicals as it provides a sharp contrast between adjacent values.
There is now an option to see both the inputs to each sensitivity run as well as the outputs in the currently displayed graph or table. See Sensitivity Parameters dialog (Model Only) for more detail.
If you click on an empty part of the model when using the ghost tool the Find dialog will come up allowing you to ghost a variable without first navigating to it.
Under the Window menu on the model and the interface, the Restore Window Layout option will reset all window, panel and tool bar positions to where they were when Stella was first installed (isee systems default settings). See Window Menu
In the Zoom control, a new option has been added to automatically adjust the zoom level of the model diagram or interface page so that all objects will be visible within the viewable space without needing to scroll. See Model Window Mode Toolbar for details.
A combined View Results and Causal Lens has been added to the Online Model Editor. See Getting Started Online.
You can now specify a model level default number format that will be applied to all variables instead of having to do this variable by variable. This is done in the Model Settings Properties Panel. The options are the same as they are for each variable, and variables can still be set individually as well.
Instead of limiting output formatting to use a $ or % you can now specify arbitrary prefixes and suffixes for output numbers. See Scales and Ranges Tab for more discussion.
You can now format time to display years, months and days and not simply decimal values. See Setting up Time Bounds for more discussion.
You can now find what graphs or tables (or interface object) a variable is used in by right clicking and selecting Find Uses on the right (Ctrl) click menu. See Context Menu for more details. You can also locate exported tables from the Export Data dialog box
The performance of published models, especially simulation speed, has been dramatically improved and will, with a modern browser, match the performance on the desktop.
It is now possible to import and export data from the interface for published models. These are both set up using a Button (Interface). When the button is clicked the user will be prompted for a file to upload or save to. Only .csv files are supported on the interface. See also notes in Creating Import and Export Files.
The new Simulation Speed control allows you to let the interface user set the speed at which the model simulates. See Simulation Speed for more details.
You can now have objects be made conditionally visible based on variable values, or the role in a multiplayer game. This allows customization of the user experience based on the state of a simulation. See Object Visibility (Interface) for more details.
Much like the progress bar in the development environment the Display Time object shows progress during a simulation and allows the user to scrub over different times in the simulation. See Display Time (Interface) for more details.
You can now add a button that will play a sound, and have a sound be played in response to a simulation event. This works the same on the desktop and for published models, though supported sound formats vary by browser. See Button (Interface) and Simulation Event Action dialog for more detail.
You can now display equations, units of measure, and any documentation provided for a variable in a Model View. This material is displayed as the mouse moves over the elements of the model. See the Model View documentation for details.
The NEGBINOMIAL (negative binomial) builtin has been added. This returns the number of failures that occurred getting to a specific number of successes for a given probability of success. See Statistical builtins for more information.
You can also specify a minimum, maximum and sampling interval so that it is easy to create truncated or sample distributions in models, and truncated distributions for sensitivity. See Statistical builtins for more discussion of truncation and sampling.
You can quote an array dimension name and element names by putting double quote around them (as in "1-2"). This capability is in place primarily to support the importation of models translated from other languages. If you do create labels with quotes the model will not work in previous versions of Stella.
The Histogram has had options added to appear as a probability density function (pdf) and cumulative density function (cdf). See Bar Graph for further discussion.
Highlighting of variables is now easier to find even in dense models with lots of color.
It is now possible to pass parameters to a simulation set up with data collection by appending them to the URL for the model. See Data Collection for details.
Note that the BINOMIAL and GEOMETRIC builtins were not behaving the way they were documented. The probability parameter was being multiplied by DT prior to execution. If you are using these with a model that has DT of 1, there will not be any change in behavior. Otherwise, multiply the probability by DT to get the same behavior. The MONTECARLO builtin will also allow you to specify a success probability that is limited to 100% per DT, not 100%*DT per DT as it previously did.
Models from versions prior to 1.6 can be opened and saved in 1.6 then reopened in previous releases and should be unchanged. As always, new features available in 1.6 will not be recognized in previous versions. Thus, for example, additional arguments to NORMAL and other random number generators will be treated as errors in earlier versions of the software.
The new time and number formatting options introduce more complexity when models are opened in previous versions of the software. By design, if a model from version 1.5 is opened and saved in 1.6 it will reopen unchanged in version 1.5. However, if you make changes to the default number or time formats (see below) in 1.6 then those changes will be lost when opened in 1.5. When reopening in 1.6, you will need to add back in the default format changes. As you add in additional number and time formatting options this becomes more complicated and it is possible to lose additional time and number formatting information. Therefore, if you are planning to work with people using prior versions of the software it may be simplest to not set the model level defaults for time and number formatting.
In Earlier VersionsVersion 1.5.2 contains defect fixes only, and no new features.
The behavior of models running in Flight Sim mode has been modified so that they respond continuously to input changes rather than momentarily pausing. See Flight Sim Mode for details.
The functionality to create an import file from the Parameter Control panel has been updated to only save values that are changed. This will allow you to easily create import files based on optimization results and also to create an import file that captures a snapshot of the interface input control settings. See Parameter Control Panel for more details.
Context sensitive help is now available in most dialogs as well as the properties panel.
Optimization allows you to find a set of parameter values (model constants) that maximize a payoff you defined in terms of other model variables. This is similar to running sensitivity that is guided to achieve specific outcomes. See Outcome Optimization for more details.
The output of graphs and tables has been enhanced to support a number of formats that can be very helpful when working with a large number of sensitivity runs.
Confidence bounds allow you to see what percentage of runs leave values in different ranges at each point in time. This allows for a clear visualization of the divergence or convergence of values over time given different assumptions on parameters and also how concentrated the values are at any point in time. See Graph Series Property Panel for details on turning these options on. The same information can be presented in tables identifying numeric values for the different percentiles (see Table Properties for details).
Histograms are created by collecting results from different runs and grouping them into buckets for different ranges. This allows you to see how many runs fall into each of the ranges. Histograms are available as an option on the Bar Graph
Correlation plots allow you to see how one variable correlates with another over a series of runs. This is done with scatter graphs by turning on comparative and checking the option to show only a single value. See Graph Series Property Panel for details.
When a model is paused, the values of all variables, stocks, flows and converters, are available at the time of the pause. On the model window these values will be updated if you make changes to any model parameters. For tables and graphs, the interface will report only stocks at the current time, but other output devices, such as numerical displays, can be set to report the precomputed values. For more information see Initialization and Pausing. The option to pause before computing flows has been removed as this is no longer necessary.
These are now accessed through the model analysis tab 
of the properties panel for the model. Use the sensitivity tab to access sensitivity setup. See
Sensitivity Specs Panel for more detail.
Clicking on the ? icon in any of the panels will now take you to the help entry specific to that panel.
You can now specify a save interval. This is useful for decreasing the storage requirements for large models and also for keeping data collection working well. See Model Settings Properties Panel for more information.
If you are using the tabbed or dropdown styles for selecting between multiple graphs you can now set a label for the tab or dropdown entry that is separate from the graph/table title. It is also possible to reorder the entries. See Pages Styles (Interface), Graph Settings Properties Panel, and Table Properties for more discussion.
You can now use keyboard shortcuts to select tools when building models. See Select Stock and so on in Keyboard Shortcuts and Key Combinations for details.
You can now rename the graph and table pads that show when the graph or table is minimized. Editing works just as it does for variable names.
You can now use the annotation button to display a variable's documentation, equations and units of measure. See Annotation Button (Interface) for more details.
The units of measure now appear in the graphical panel - making it easier to there are units issues with graphical functions. Also, if the equation for a graphical has an error the panel will open on the equation instead of the graph.
When setting interface page size, existing content may restrict how much it can be adjusted. If a page has content that would be outside the page on resizing you will not get a message indicating the reason the page size was not able to be changed as requested.
A new Loop Label object has been added to the mode editing toolbar. This allows you to label loops and will highlight the links in the loop when the mouse passes over it. See Loop Labels for more information.
A new navigation button has been added to the model to allow position changes to predefined locations or variables. These can be useful as shortcuts for navigating through large models. See Navigation Button (Model) for more details.
Use NAN to set a variable value so it will not appear on graphs. Use CLOCKTIME to return the current time on the computer the software is running on.
There is now an option to fit the graphics content to the frame. See Graphics Properties for more detail.
In addition to publishing model interfaces, you can now create multiplayer games in which players connect from their own browsers, make decisions, and indicate their readiness to proceed. The model then uses decisions from all of the players as inputs for the next step in the game. Multiplayer games can be published to the isee Exchange under the Creative Commons license, with some limitations on model size. See Multiplayer Games for more details.
It is now possible to create sophisticated animations based on SVG graphics files. These animations can include color, size, and rotation. See Animation Object (Interface) for more details.
When you add a button to your interface, you can now set conditions under which each of the actions will be executed. This is especially useful in multiplayer games. See Button (Interface) for more detail.
Published interfaces can now be made resizable, allowing them to work well on a variety of devices. For details, see Resizable Interfaces.
You can now collect data on the use of your published interfaces, including the results from different users and the time spent on different pages. Data collection is available through an isee Exchange subscription. There are several reporting options available for analyzing the data. For more details, see Data Collection.
The modeling language has been extended to allow the use of expressions inside of the array functions (such as SUM(a*b)) to take an inner product. An inner product operator has also been added for 1 and 2 dimensional arrays. See Working with Arrays and especially Specialized Array Manipulation for more discussion.
There are new options for navigating through multiple pages of graphs, tables, and input devices. Rather than the standard 
and 
buttons, you can use tabs for each page, or a dropdown list including all pages. This is applicable to graphs, tables, group inputs, and arrayed inputs, and is selectable from the properties panel for each. See
Pages Styles (Interface) for details.
When you connect a variable to a module, or a module to a variable, you will see a menu item allowing you to select which variable within the module to connect. This will appear only if there is something marked as an input or output within the module that has not yet been connected.
We have made a behavior change when you are using a separate initialization equation for a graphical function variable. Previously, the I equation would determine the value of the variable. Now the I equation determines the input into the graphical function.
Graphs will now use the initial values computed in the background. For bar graphs and pie charts, this means they will be populated with initial values before and after running an output restore. Regular graphs will not draw, but they will have a scale. This is true for models and interfaces, including published interfaces. Note that changing a controlled value will no longer generate a new initialization. See Known Issues for more discussion.
We have adjusted Stella Live to be less sluggish on larger models, and also to give visual cues when the model is in the process of computing. In particular, as you dial a knob, while the software is computing the graphs, over model elements will be displayed in gray instead of red. This indicates that the graphs are potentially out of date relative to the value showing for the knob.
If you try to make a connection that would cause a circularity the list of variable involved in the circularity is not displayed in a dialog box.
Hovering on a connecter into or out of a module will now show which variables are being connected.
You can now navigate out of modules using keys. To get to the containing module use Ctrl+Up Arrow (⌘+Up Arrow on the Mac). To get to the home module use Shift+Ctrl+Up Arrow (Shift+⌘+Up Arrow on the Mac). See Keyboard Shortcuts and Key Combinations for other shortcuts.
The three line limitation on variable names has been removed.
You can now add polarity symbols to flows.
On the interface, the auto-alignment grid now includes objects in templates so that you can more easily manage page layout.
Keyboard scrolling using the arrow keys has been improved.
Array names and elements have been added to the auto-complete when working with arrayed equations.
The scales and ranges tab has a new icon (
) that is more representative of its function.
We have substantially improved the performance of published models of significant size.
You can now edit and work with your models online by logging in through the isee Exchange. See Stella Online for more details.
Available for both the model and interface, we've added two new options for displaying results. The Pie Chart is a new output format for the display of variable values at a point in time. The Area Graph allows you to display a series of graph values stacked on one another, with distance between them based on the values of the variables. Both of these are configurations of graph objects and can be selected in the Graph Series Property Panel.
The new Allocation Input (Interface) object allows users to set values for inputs that need to add up to a set amount. It's implemented as a draggable pie chart, with adjacency determining the sequence in which other inputs change as the user makes adjustments.
A reset button (
) has been added to interface input objects, allowing them to be reset to the model value by the user (the button also marks the control as changed). The option to use this is set in the
Interface Settings Properties Panel.
Buttons on the interface can now execute either an import or an export of an on demand data link. These actions do not do anything for published models. See Button (Interface) for more details.
Graphical input objects on the interface can now be changed either by tracing the shape, or by switching the Table view and entering xy pairs. See Graphical input (Interface) for more discussion.
You can now import both the x and y axis for graphicals by putting :x and :y after the name of the graphical in successive rows or columns. See Creating Import and Export Files for details. The import options now handle arrayed inputs more consistently and better manage compatibility between US and European number formatting.
The Parameter Control Panel now highlights variables that have been overridden on the panel, making it clearer what has been changed (this is especially important for values already imported).
Any graphical functions that use a constant as an input will now be marked as invalid. This makes it easier to uncover issues where TIME might have been the intended input. If you do want a constant as an input, you can use an expression (as in 0+0) to prevent the error.
The INIT and ENDVAL builtins can now take an expression in addition to using a variable directly.
You can now control the colors and line styles for the different lines appearing in comparative graphs. This is set for each model (and there is a default setting for new models).
Window positions are now preserved for all dialog boxes and panels, making it easier to get back to the same setup you've been working with.
By right clicking on a graph, table, or numeric display (or any interface control with a variable in it), you can jump to the definition of that variable in the model (not available for Stella Designer).
You can now format multiple items on a graph or table using the right click menu.
You can now preserve graphs and tables with a single click so that they won't change with new runs or be cleared when you restore outputs. This is done from the Graph Series Property Panel.
We've added the functionality 'import runs from external data sources' to the Data Manager. This makes it easy to compare results with measured data or with results from other models.
A number of other improvements have been made to the Data Manager, giving you more control over the way in which runs roll off the list and what runs are displayed in graphs and tables.
The import and export dialogs have been improved to make it both easier and clearer to set up imports and exports. The use of temporary imports has been reintegrated into the import dialog so that everything is in one place.
The parameter control window has been enhanced so you can see the original model value, and to make it easier to reset and control different variables.
Any variables appearing in the parameter control window are also marked on the model diagram (
).
The simulation speed slider has been replaced with a field that allows you to input how many seconds you would like the simulation to take.
A speedometer style gauge has been added to the interface to allow you more variety in presenting results.
Spatial map has been integrated into the interface, making it easy to present spatial information in an appealing manner.
An Array input device has been added to allow sequences of array values to be input into the interface.
Explore mode on the interface has been replaced with Experiment Mode, which behaves the same way as Presentation mode does, making it easier to test your interface.
There is now a button action to reset a variable to its equation (or imported) value.
The switch object can now be configured to appear as a checkbox.
You can now optionally animate diagrams shown in the Model View, including those shown for storytelling.
The Model View controls have been improved to make it easier to include model elements on interface pages and stories.
The Pages Panel (Interface) can now be collapsed to allow you more room for editing.
Before running, control changes will update model values including numeric displays, speedometers, and spatial maps, allowing users to see the initial conditions that will be used as they set things up.
Sketchable graphs update automatically in Stella Live.
Stella Live executes more quickly during development and on published interfaces.
Embedded images display more quickly.
The thumbnail drawing is much faster.
Divide by 0 errors are now reported on the interface.
Models are continuously reinitialized as they're built so that you can see the effects of equation changes immediately by hovering over a variable.
There is also an option to perform a new simulation each time you change model structure, which allows you to see the effects of changing connections and equations immediately.
Circularity checking and reporting have been improved and now detect subtle initialization dependencies related to leakages and discrete stock equilibrium initializations.
Circularities are now reported one loop at a time, and all the variables in the loop are marked invalid, making it easier to visual the loop.
Non-stock equations can now specify an initial value to be used when the model stocks are initialized. This is very helpful for breaking initialization circularities that have been identified.
We have added a // operator which works the same as SAFEDIV, giving 0 instead of an error when dividing by 0.
Divide by zeros are more accurately reported, even when they occur within expressions such as MIN. Legacy models will automatically have some of this reporting masked to maintain compatibility.
Checking Unit Consistency now reports where in equations unit inconsistencies are detected.
For locations that use a , as a decimal separator, you can now enter equations using this, as well as ; as an argument separator. This can be turned on and off in the Preferences Dialog.
The registration dialog behavior has been changed so that clicking twice on the button to open a web page does not cause two or more activation requests to be sent to our server.
Sliders and Numeric Inputs now allow you to specify the width of the slider or input box in order to get better horizontal alignment when placing one object below another.
Sliders, Numeric Inputs, Knobs and Switches on the interface now have an option to wrap text.
The Causal Lens now highlights the path of selections you have made to make it easier to see the sequence of causality.
The interface development tools are all new. The new interface development and deployment capabilities take you from design to presentation and, optionally, the web, quickly and cleanly. See Building Interfaces for more detail.
Interfaces developed in previous versions of the software will be translated, and most of their functionality retained, though some adjustment will be necessary to get optimal results. See the notes on Using Version 10 Models for more comments on interface translation.
Version 1.1 adds a number of new and updated panels that can be used to analyze and navigate your model.
The Causal Lens™ displays graphs for a variable and the inputs to that variable, helping you to better understand the sources of model behavior.
The Results Panel now includes both a graph and a table, allowing you to easily look at specific values at specific times. There's also a checkbox to not round the scales, which is helpful when looking at variables with limited movement through a simulation.
The Message Log records any variables that go above or below their specified ranges, as well as graphicals with inputs above or below their specified minimum and maximum x values. The same log is also used to report messages about anything that happens when opening a model.
The Parameter Control Panel allows you to see what variables are changed relative to their model equations, and also make changes to those variables.
The Selection History - Obsolete panel keeps track of which variables you've selected, making it easier to go back and see where you've been. Click on an entry to take you back to that variable.
We've improved selection highlighting to make it easier to see which variables have been highlighted. This is especially useful as we've added a number of new ways to navigate between variables.
You can right-click on a variable in an equation and jump to that variable. There's also a Causes/Uses menu to navigate to variables that cause, or are used by, the variable you're looking at. See the discussion on the Equation Tab of the Properties panel for more details.
We've improved the error message reporting in equations, and made this reporting somewhat stricter, especially around the MIN and MAX functions, to prevent ambiguity.
The units messages now tell you how many variables you have left to define units for. If you haven't defined units for any variable, the dropdown simply reads Define Units.
There are no longer any restrictions on what names you use for variables, except that they must be unique and not conflict with any Builtins. When used in equations, variables that have operators or other characters will be quoted and escaped. (If you type them in, you'll need to do this yourself, but it happens automatically if you choose them or use autocomplete.) Variables with a period (.) in them will also be quoted when displayed on the diagram (this is to prevent confusion with module qualified names).
Models open significantly more quickly than before, letting you get on with development. Exporting of data is significantly faster and will happen in parallel during simulations.
An analog to the SUM function takes the product across elements of an array.
An option has been added to the Preferences Dialog that turns off Stella Live when in Explore mode. This allows you to work with larger models and make changes on the model window, but only simulate them when you explicitly request a run.
You no longer need to go to the home module to change run specs or other model wide settings. These are available in the panel from any module.
You can now order equations by sector, and also view values in the Equation Viewer. Complete information about model size has also been added to the end of the equation listing
We've redone what used to be called Posters or Simulation Messages into an easier to use system of Simulation Events that can be applied separately to the model and interface. Existing models will have these converted to the new format when opened with the new products. Navigation events will need to be redone - see Simulation Events.
We've added new export formats to export a single value or set of values at a time, and improved the way tables are specified for exporting - see Export Data dialog box
Version 1.0.3 is designed to work well with high resolution displays, including Retina displays on Mac. The icons on the screen will automatically be enlarged to accommodate displays with high pixel counts, and the diagrams will be presented so that they're easy to read at default zoom levels.
We've updated the icons used throughout to be clearer and more consistent, in addition to working better on different resolution monitors.
The movement back and forth between Edit and Explore modes has been streamlined. Changes you make to constants in Explore mode, by adjusting knobs or entering values in the Results Panel, will persist if you enter Edit Mode and then go back to Explore Mode. Modifying an equation will reset any value you set in explore mode, but only for that equation.
You can set values for arrayed constants in the Results Panel in Explore mode (arrayed constants don't have adjustable knobs). Just open the results panel and click on the constant you want to change. You can select the array entry to change from the dropdown at the top of the results panel.
Constants now have an increment setting that can be used to control the number of values they'll take on when changed via a knob in Explore Mode. This can be very helpful if you want to limit the number of values that a constant can take on, and will even allow for the addition of switches that take on only 0 and 1 as values. The increment is set in the Scales and Ranges panel.
This functionality has been moved to the Panel to make it easier to use. Rather than opening the import dialog, open the Properties Panel on the model and there's a new tab 
, to make temporary changes, including importing interface settings from STELLA and iThink Version 10.1.2 models.
Clicking on the add variable button in the graph and table panels will now open up the Find dialog, making it easier to select and add variables.
Copying model diagrams into the clipboard is now more sensitive to context, and will remove the borders from graphs when they're copied by themselves. This makes it easier to paste your content into other applications.
Now when opening a saved model, it will open in the same view in which it was saved.
Substantial improvements have been made to Stella Professional's computation engine. Not only will models open more quickly, but large models should simulate three to eight times faster than in the previous version. This allows you to get more done more quickly.
A new function, SAFEDIV(<numerator>, <denominator>, [<onzero>]), has been added, which allows division in cases where the denominator might become 0. When this happens, the function will return 0, or the optionally specified third argument. Use this to prevent simulations stopping because of a divide-by-zero condition.
The Import functionality has been enhanced to allow you to import values without making any changes to model structure. This is useful for running experiments in which you don't want to change the model, but just see the results of changing constants and graphical functions. This also allows you to override computed variables with a constant value, much like you would when, for example, attaching a slider to a converter on an interface.
An option has also been added to the Import dialog that will read interface control values from a model created in version 10.1, and create a temporary change file. This allows you to simulate models in Stella Professional and get the same results that they generated in the Version 10 products.
The Alt key now works by holding it down, rather than tapping it, when you want to repeat a build step, such as drawing connectors between variables.
When you position the mouse over a connector handle (or arrowhead), the color of the connector will change, to indicate that it's active and can be dragged. Similarly, when you position the mouse over a variable name, the name will be highlighted with a box to indicate it can be repositioned. This makes it much easier to adjust model layout, especially for densely connected models.
A Windows menu has been added to allow you to more easily navigate between models when you have multiple models open. Under the Windows menu is an item to open the Results Panel. A Run Specs menu item has been added to the Model menu, as an alternative way of opening the Run Specs panel.
Registration information can now be accessed and changed through the Registration dialog box opened from the Help menu. This allows you to change your registration information and unregister the software from one computer, so you may use it on another.
When you place a connector, the style panel will contain an option for a Delay Mark. See Styles Tab for Building Blocks for more details. The polarity markers can be shown using either +/- or s/o, as specified in the Model Settings Properties Panel
Set this on the Model Settings Properties Panel to shorten the names displayed in graphs and tables (e.g., have profit be displayed instead of accounting.profit).
You can now create a Sketchable graph for any graphical function in the model, without needing to add another variable for comparison. This allows you to use the Sketchable graph as an input control. See Sketchables for more information.