A payoff is a single number that summarizes a simulation. You need to define a payoff before you can run an optimization or do Kalman filtering. You can also define a payoff for normal simulations. At the end of the simulation the value of the payoff will be reported and added to the error history window.
The payoff for a model can be defined in terms of a comparison of model variables with actual data, or as a combination of model variables. These two types of payoffs are known as calibration payoffs and policy payoffs. The payoff is defined inside of a Vensim Payoff Definition (.vpd) file. This file is referenced as second step the Optimization wizard takes or is named in the Payoff Definition field on the Advanced tab in the Simulation Control dialog (see Chapter 8).
If you are using the Optimization wizard the Payoff Definition dialog will appear in the second step. From the Simulation Control dialog click on the Ed... button next to the Payoff Definition field.
Filename is the name of the Payoff Definition file. If no filename has previously been specified Vensim will use the model name or, when using the Simulation Control dialog query you for a name. You can type in a different name if you like. If you do this and that file already exists you will be asked if you want to overwrite it. You do not need to include an extension, Vensim will automatically add one.
Choose New File... allows you to choose a different Payoff Definition file to work with. When you click on this a File Selection dialog will open. Select the file you want to work with and its contents will replace those in the dialog.
Clear Settings will remove all the parameters and reset the Type to Calibration.
Type: Calibration, if checked, will attempt to match model variables to data for those variables. A calibration payoff requires that a data file be specified. If Policy is checked, the payoff will be computed based solely on model variables so a data file is not necessary to compute the payoff, though one may be necessary to simulate the model.
Payoff Elements lists the different components of the payoff. For calibration payoffs, this is a list of the model variables that should match data. The list takes the form:
model variable|data variable/weight
Variable is the name of a model variable (model variable) that you want to compare to available data.
Compare to is the name of a variable (data variable ) in a data file, or a computed Data variable in the model (defined using a := equation).
Weight is the amount of weight you want to give to the error between the variable and what you are comparing it to. See the discussion of payoff computation below for a more complete description of what weight does.
If you follow the recommended convention of using the same name for data and model variables you can use the simpler form:
model variable/weight
For Policy payoffs this is a list of variables contributing to the payoff. The list takes the form:
model variable/weight
which is identical to the simple payoff form for policy optimization except that weight has a different interpretation. This is discussed in more detail in "Payoff Computation" below.
NOTE The weight must be separated from the variable name (and any data name) by a slash /. Comments in braces {} placed anywhere on a line are ignored. If the specified weight is not numerical, it must be a model variable (usually a Constant, but Auxiliary, Level, Data, and Initial are also valid types here). If no weight is specified, an error will occur.
Delete Selected deletes the selected Payoff Element.
Modify Selected brings the selected Payoff Element into the editing boxes below.
Add Editing adds the contents of the editing boxes into the list.
Variable is the model variable to include in the payoff. You can type in a variable name here or click on the Sel button to select a model variable.
NOTE If you are using a model with subscripts, the full set of subscripts (with Subscript Constants not Ranges) must be specified here.
Compare to is the name of the data series with which to compare to the model variable. If you use the same name for the data series and model variable (recommended) you should leave this blank. For a Policy payoff definition this is always blank.
Weight is the weight to be given to this part of the payoff. This can be a number or a constant in the model. It should be positive for calibration payoffs, and can be negative or positive for Policy payoffs. See Payoff Computation below for more details.