To forecast a negotiation, specify for each player only four variables: the policy position they publicly advocate, how much they care about the issue (salience), how much clout they could bring to bear (potential influence), and how flexible they are willing to be. Modeling these interacting choices, rather than consulting experts for a single verdict, generates a prediction of the outcome.
People are assumed to act to advance their own perceived interests rationally, so their behavior becomes predictable once you know what they want and what they can do. The model does not require people to be nice, moral, or even correct about the world, only to pursue their own goals as they see them.
The person formally in charge is frequently not the one who determines an outcome; real leverage sits with whoever holds salience-weighted influence over the specific decision. Map the actual network of stakeholders issue by issue rather than assuming authority follows the organizational chart.
Stating predictions precisely and in advance, with specifics that can later be checked, is the discipline that separates genuine forecasting from retrospective storytelling. Being willing to risk public embarrassment by committing to a falsifiable claim is what makes a method testable and improvable.
The car salesman sets price by exploiting the buyer's information disadvantage, so the buyer should neutralize it: get competing dealers to bid against each other by phone for a precisely specified car, commit to buying from the lowest quote, and remove the salesman's room to maneuver. Changing the game's information structure, not haggling harder inside the given game, shifts the outcome.