NOTICE: I have recently learned that it is possible for email addresses of STUDENT and FACULTY users to be accessed by others besides me after you have registered. I cannot currently correct this so when you register be aware that this is the case.
I am removing access to the apprentice version of the game. You will still be able to enter data for it but will be asked to log in with an appropriate level of registration to actually run the game. Registration is available ONLY for students and faculty engaged in a bona fide university course (with appropriate communication to me to that effect and my approval) or research or to professional-level users. The professional level of access is only available for a fee which needs to be arranged in advance so please do not attempt to log in to either of these levels if you do not meet the criteria. Approved users, click below to access the Predictioneer’s Game. Have fun Predictioneering!
By playing the game, you agree to the END USER LICENSE AGREEMENT
USERS: PLEASE NOTE that due to a recent upgrade of the software there have been some changes in the format for input data. To enter a data set successfully using an excel .txt file you need to have a first column in the data set called group. It can be otherwise empty; that is, it does not need to have any data in it. That is followed, as before, by data on player name (one word label such as player or stakeholder and each name must also be without spaces or punctuation as before)), influence, position, salience, flexibility and veto (a zero for non-veto players and a 1 for any veto players). After that you need as well to have a column that can be labeled fixed, another called random and yet another called optimize. The columns for which you don’t have data (fixed, random, optimize) should contain 0 (zero) in each player’s cell.
The issue scale and data on Zimbabwe were constructed by undergraduates in a course called Solving Foreign Crises offered jointly by the Alexander Hamilton Center for Political Economy and the Department of Politics at New York University. The issue data set for it were constructed in late January or early February 2009. I thank my students for permission to use their data. The data can be found below and provides a template for constructing data sets. Remember, input data files MUST be in .txt format. The issue scale is defined here:
Issue: Power sharing:
Which government is most viable?
0 No Mugabe at all. Mugabe is completely removed from the government.
10 Powerless Mugabe. Mugabe is a figurehead.
25 Involved Mugabe. Mugabe holds a minor government position.
40-45 Power sharing, Mugabe weakened.
50 Equal power sharing
65-70 Power sharing, Mugabe is the most powerful player in government.
85 Mugabe is President and mostly in charge but makes concessions on cabinet membership.
95 Mugabe is President, keeps opposition out of cabinet but allows some opposition in
100 Mugabe is the sole dictator in Zimbabwe.
Zimbabwe Data Set:
The Predicitoneer’s Game© Software Training Manual
By Bruce Bueno de Mesquita
Welcome to the Predictioneer’s Game Apprentice Software©. The following guidelines will allow you to correctly enter issue-specific data into the program and then interpret the model’s results. The quality of the information you provide is critical to the reliability of the analysis, so please follow these directions closely.
The issue analysis Apprentice Version you are using is for instructional purposes only. It produces exactly the same forecasts as the Professional Version but limits the information available to you to restrict opportunities to engineer outcomes. Although some engineering will still be possible by altering initial positions or salience or resolve, it seems best to restrict the range of ways that events can be engineered so that apprentice users do not unwittingly influence events in unintended ways.
The meaning of each variable is explained below. After you master the meaning of the variables you can begin to enter data directly into the program.
The first data column identifies the name of each player, whether the player is a group or individual. Names must not contain spaces. It does not matter whether you use all lower case, upper case, or a mix. Only the first column is alphabetic. Everything else must be numeric.
The second column contains data on the relative potential influence of each player. This must be greater than 0.
The third column shows a numeric value for the outcome on the issue that is currently advocated or supported by each player. The range of values will depend on how you defined the issue.
The fourth column of data assesses the salience each player attaches to the issue. This will be values greater than 0 and less than 100.
The fifth column assesses a player’s resolve or flexibility. It takes values between 0 and 100, including the possibility of a value of 0 or of 100.
An optional six column consists of 0s and 1s with 1s being assigned to any player who has the right to veto an otherwise agreed-upon outcome. Do not confuse veto authority with influence. It is possible to have low influence and still have a veto. Lawyers, for instance, or financial advisers, often have a veto over a negotiated agreement between firms planning a merger even if they are not very influential.
Definition of Issue: An issue is any specific policy question for which different individuals, organized groups, or informal, interested parties (that is, stakeholders, actors, players) have preferences regarding the outcome. The range of preferred outcomes on an issue must be capable of being represented along a single line or continuum. Be sure to define carefully the precise policy question you want to analyze. The right and left ends of each policy continuum should specify the most extreme outcomes actually supported by any interested party. Of course, these extreme outcomes need not refer to a resolution that anybody believes will be achieved, but refer only to the fact that there is at least one stakeholder that currently seems to support such an outcome.
You should associate additional labels with other points on the continuum to reflect a progression on the line from (a) the most radical resolution supported by anyone; to (b) progressively less radical resolutions; to (c) progressively more conservative; to (d) the most conservative resolution supported by anyone (or from most conservative to most radical). Additionally, where appropriate, it is useful to identify the location of the status quo on the issue continuum. Not every issue has a status quo, but it is useful to keep a record if it exists.
When a status quo position is identified it is important to reflect on how it was chosen and how the issue continuum is defined relative to the location of the status quo. Sometimes the status quo is at one or the other extreme end of the continuum, but this is uncommon. More often, when there is a status quo it is located somewhere between the extremes of the issue continuum. Sometimes locating the status quo at one or another extreme could bias the analysis by assuming that no one supports a position farther out than the status quo. That is an uncommon situation. Just because some policy is the status quo does not mean that some interested players would not like to change the policy in one direction or the other.
In defining the meaning of points along the issue continuum, it is useful to keep in mind that these defined points are part of an assumed continuum. They can represent discrete outcomes in a discrete space but more often the issue is likely to have continuous outcome possibilities. The spacing of defined points should reflect their relative substantive distance from each other. That means that you should avoid just dividing the issue continuum into equally spaced points and defining those without regard to their relative substantive proximity.
Below we will see how to turn verbally described positions into numeric values.
Definitions of Variables
The essential prerequisite behind the development of the inputs is to begin with precise definitions of each variable.
Enter the name of each stakeholder in the first data entry column, one stakeholder per row. Do not use more than one word to name a stakeholder; that is, do not put a space between words. For instance, if you have a stakeholder named John Doe enter it as JohnDoe. The length of the name does not matter as long as there are no spaces. You can use dashes, apostrophes and so forth.
A stakeholder or player is any individual or group with an interest in trying to influence the outcome on the issue being analyzed. Do not limit the list of players just to those who will ultimately make the decision if others care to weigh in, trying to influence the decision makers. All who try to influence the outcome should be represented in the stakeholder list.
Stakeholders can be treated as a single group or player provided that their position, salience, and resolve/flexibility score is the same even if their influence value is different. When a group is constructed out of many individuals, it is fine to sum the influence of each into the group’s total. But remember, a group must consist of interested parties whose value on the remaining variables is the same.
The value assigned to each player in this column reflects the relative potential ability of each player to persuade other stakeholders to adjust their approach to the issue to be more in line with the influencer’s perspective. The values typically will fall between 0 and 100 but they are not restricted to this range. They can be larger or smaller than 100 but must be above 0. Here are some useful rules of thumb for getting started in estimating player influence.
100: The most potentially persuasive stakeholder on this issue. There can be more than one group at this score or at any other score. The value 100 is illustrative and is convenient to use, but the software does not restrict resource estimates to be between 0 and 100, the values can be larger than 100.
The resource variable evaluates the potential ability each player has to persuade other stakeholders to support a point of view on the issue in question. The ability to persuade may be derived from holding a position of authority, being an expert, commanding a large budget, or any other factor that makes others listen to someone. Below is a useful way to think about constructing all other player potential influence values relative to the most influential player’s score of 100.
All other values:
A stakeholder’s value must be positive and must be evaluated relative to 100 (or the maximum score assigned) and relative to the resource values for other stakeholders. So, two stakeholders with 40 and 60 would equal the one stakeholder at 100 in a head to head contest with no one else involved if each of these three stakeholders tried as hard as they can. Two groups at 15 and 30 would, if they shared a common position, be very close in potential influence to a group at 40 and probably would just barely persuade the 40 to accept their point of view if there were no other players involved.
The influence scores should not be thought of as percentages. A decision maker with a score of 100 does not have 100 percent of the potential influence and may, in fact, have only a small percentage of the total. The total, of course, is the sum of all of the potential influence across all of the groups or decision makers.
The position preferred by each stakeholder on the issue, taking constraints into account. This position is not likely to be the outcome the stakeholder expects or is prepared to accept, nor is it likely to be what the player wants in his or her heart of hearts. It is the position the stakeholder favors or advocates within the context of the situation. When a player’s position has not been articulated, it is best thought of as the answer to the following mind experiment: If the stakeholder were asked to write down his or her position, without knowing the values being written down by other stakeholders, what would he or she write down as the position he or she prefers on the issue continuum? To place a numeric value on the position, the investigator must first have defined the issue continuum. The continuum will either have a natural numeric interpretation, such as the percentage of uninsured on health care to be covered under a new policy or the analyst will need to develop numeric values that reflect the relative degree of difference across policy stances that are not inherently quantitative. It is important that the numerical values assigned to different positions (and they can range between any values) reflect the relative distance or proximity of the different solutions to one another. An easy way to turn player preferences on an issue into numeric values is to place each player on the issue continuum you defined, locating then at the point on the continuum that reflects the policy they support. Then, use a ruler to measure how far each player is from one end of the line that reflects the range of choices. Let the left-hand end of the line equal 0. Then each other point on the line is simply its distance from 0 on the ruler.
Salience assesses how focused a stakeholder is on the issue. Its value is best thought of in terms of how prepared the stakeholder is to work on the issue when it comes up rather than some other issue on his or her plate. Would the stakeholder drop everything else to deal with the issue? Would the player work on it on a weekend day, come back from vacation, etc.? The more confidently it can be said that this issue takes priority over other matters in the stakeholder’s professional life (or personal life if the issue is about a personal or family matter), the higher the salience value.
90-100: This is my most important issue. I would drop whatever I am doing and turn to this issue whenever asked.
70-80: This issue is very important to me. It is certainly one of my most important issues. I would try very hard to reschedule to handle this issue when it arises.
50-60: This is one of several important issues. Others are more important. I would have to drop this if one of those others arose, but otherwise I will try to focus on this issue.
30-40: This is an issue I care about, but it is not that important to me. I have many more important issues to deal with and so generally would not drop what I am doing to deal with this and generally would focus on something else.
10-20: This is a minor issue to me. I rarely pay attention or make much effort.
Less than 10: I really don’t care about this issue.
Every stakeholder is assumed to care about two dimensions when addressing an issue. The position variable, discussed above, assesses the outcome the player currently advocates. Flexibility/Resolve evaluates the stakeholder’s preference for reaching an agreement as compared to sticking to his or her preferred position even if it means failing to reach an agreement. The variable ranges between 0 and 100. Higher values reflect greater flexibility; lower values greater resolve. The meaning of alternative values is illustrated below.
90-100: Overwhelmingly prefers reaching an agreement and being a party to it. The stakeholder is prepared to accept almost any outcome on the continuum if it means resolving the issue.
70-80: Reaching an agreement is considerably more preferable than showing resolve and sticking to one’s position, but the stakeholder has limits concerning how far s/he will go on the continuum to make a deal.
50-60: The stakeholder has a fair amount of flexibility regarding the outcome, but is mindful of trying to promote seriously the position s/he prefers. Reaching agreement is about as important as promoting an outcome favored by the stakeholder. Few players are routinely much higher than this to start with. Of course, some are so take this observation as just a rule of thumb.
30-40: Reaching an agreement is considerably less preferable than showing resolve and sticking to one’s position, but the stakeholder is open to significant concessions on the issue dimension in order to improve his or her welfare on the flexibility/resolve dimension.
10-20: The stakeholder strongly values the position s/he has advocated although s/he will make some significant concessions to reach an agreement not too far from his/her current position. Losing is preferred to being a party to a deal that is not close to the stakeholder’s preferred position.
Near 0: The stakeholder is almost completely intransigent so that there are very few issue resolutions s/he will agree to and they must be very near the stakeholder’s preferred position. The player is highly resolved and prepared to lose rather than offer more than minor concessions.
After you have run the model – I will explain how to do that below – you will see that an output file has been generated with a name you gave it. The output file is in a .csv format (comma separated variables) and can be read in Excel or many other spreadsheets. The output consists of 4 elements: Round-by-round player positions. The first round is, of course, the positions you entered. Subsequent rounds are the values calculated by the Predicitoneer’s Game©. Next is a Forecast Matrix. This includes the issue forecast, the round-by-round forecast, the security forecast, values based on two alternative rules for predicting when the game will end, the upper and lower bound positions of any veto players, and a report on whether game has not ended (0) or has ended based on either end-game rule (1) or based on both rules (2). I judge the game to end in the round in which a 1 appears for the first time in the end-game row of the Forecast matrix but this is not a hard-and-fast rule. You may wish to evaluate the likely forecast in the round or two immediately before and after then end of game rule is met to judge how confident you are in the predicted value.
The issue forecast takes surround round-by-round predictions into account. The round-by-round predictions are equal to the weighted mean value of the positions of all of the players in that round. The security forecast is the weighted median position of all of the players in the round. The weighted mean or its smoothed version (the issue forecast – my preferred basis for prediction) is the reliable basis for predicting since the player’s value for an issue is taken to be two-dimensional (a weighted combination of their policy stance and their eagerness to reach agreement or resist agreement).
The output then turns to the round-by-round summary of actor relationships. The values are percentages. No Dispute calculates the percentage of pairs of players who share the same position. Status Quo is the percentage of players whose relationship with each other is to leave each other alone. Compromise refers to the percentage of player interactions in the round that are predicted to involve their compromising on an intermediate position somewhere between their current stances. Coerce refers to the unilateral imposition of costs by one player on another who then gives in rather than resist or to a player who gives in to another in anticipation of the other imposing unacceptable costs if they don’t give in. Clash refers to relationships in which each player in a pair imposes costs on the other.
The output then reports the members of the pivotal coalition of players in each round. These are the players who, if they materially shift positions, will leave insufficient support for the outcome anticipated in the round to be sustained.
Finally, the output shows the power of each player in each round. Power is a combination of changed values for each player’s influence and each player’s salience for the issue.
The Professional Version of The Predicitoneer’s Game© provides much more detailed output in this first file of results and also provides 3 other files with details about interactions, costs and benefits for each pair of player in each round as well as other pertinent information about beliefs and coalition networks.
Options in Software:
After entering data, you can choose the number of rounds to allow the model to run. You can also elect to randomly shock any or all of the input variables round by round. For instance, if you click on the Shock Salience box and enter a 30 in the Salience Shock Probability box, then in each round each player will have a 30 percent chance of having its salience increased to any value – randomly chosen by the program — in the admissible range. Repeating analyses enough times with such shocks helps identify how robust or stable the predictions are based on the initial data.